0:00Democrats, right? Why does this matter?
0:02Why does this matter? Because in midterms this century, take a look here. The Dems win the House 2000 midterms. Let him party ID. They won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
0:15They never won the House. So, what we see here is because Democrats leading on party ID, that says that they're going to win the House. We are talking about record high spikes in gas prices. I mean, just look at this. Okay, you know, normally gas prices do go up during the course of an election year, during the course of a midterm year, but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
0:42for a midterm cycle this century is 18%.
0:45But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
0:52Look at that. My goodness gracious. By far the highest on record. prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps and this is for the year this election season if you're thinking about heading past Labor Day into November. Where does that compare to past years?
1:10Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year, however, the gas prices have actually been going up up and away.
1:28They have actually risen up 6% over the last month. So, normally you're getting, hey, there's some some relief going on with gas gas prices as we go into the midterm election. Not this year. In fact, the relief not there at all. In fact, there is more pain at the pump now.
1:44Every other election, every other election, midterm election this century at this point, what you're looking at, the prior month changed, the gas prices all went down. This year we're seeing the exact opposite. Gas prices are going up.
1:56All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
2:01Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. Willing to pay more because of the Iran war.
2:11Overall, 71% say they unwilling compared to just 29% who say willing. Look at this number among independents. 81% 81% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
2:26What are the prediction market saying about where gas prices might be by election day?
2:29Yeah. Okay. You might say, "Hey, maybe able to be some relief. No more $4 gas prices." Uh uh cow she prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big big league problems for Republicans running. These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He
2:55had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos. Goodbye. See you later. Look at his net approval rating now in an average of polls CNN SSRS polls across 2026. He's 42 points underwater. That is a switcheroo Kate
3:23Baldwin of what? 43 points shifting away from the president of United States. His support rapidly declining, plummeting among workingclass Latinos, a group that helped get him elected back in 2024.
3:38Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue, but one issue on which in the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest name see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are, of course, the 435 members of Congress.
3:59Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lorie Tran, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
4:28Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
4:57capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Amade now says he is going to start doing and wants the others to do as well.
5:18Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that's just that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
5:48there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
5:59Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened. Yeah. And and and Amade comes out with this proposal on on Saturday that you're no members of Congress coming back until after the afterction the elections.
6:20That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
6:37Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China. This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate."
6:58What do you say to that argument?
7:00I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards and I'm worried about
7:28entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
7:40Uh, Congressman, I really appreciate it.
7:42Thanks so much for your time.
7:44Lori Tran, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
7:52Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
8:01Anything's headed? Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
8:19Um but you know there has to it has to result in real legislation real global uh cooperation especially you know there's this upcoming visit by she but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
8:39I want to play you something that um you all know Harari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
8:49You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
9:16this idea that yeah these companies are creating kind of AI superhum super beings or gods and expecting them to cowtow to us seems I mean he makes a good point I think crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and
9:45anyone who understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials. And I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
10:14something that's so important if it's as important as they say it's too important for them to run very much like nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
10:43Garrett Swisser, thanks so much.
10:47The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
11:15right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
11:33Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
11:38And why is that? Why is that a good idea?
11:41Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
11:56If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
12:09Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing, they throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
12:32I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
12:50Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence, a
13:19pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails, so to speak.
13:29What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
13:38I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
14:05model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
14:30That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
14:45So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
14:57Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily and they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it?
15:20They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing AI from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time?
15:48You said something earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is right. It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
16:20If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
16:49we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
17:09Would you say that you you're an optimist when it comes to these issues?
17:13Um, no. Um, I'm hopeful.
17:17But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
17:46stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
18:03So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
18:25Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
18:54been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the Hugging Face hack where OpenAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
19:23that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
19:53But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
20:21are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent and who's going where. There's a small pool of people to choose from.
20:36These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
21:05is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
21:27President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
21:36The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's not approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result.
21:59How about among independents? You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
22:24Okay, we're going to go from the galaxy back to here on Earth.
22:28We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
22:40Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
22:56I mean, just look here. Okay? Local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
23:25we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
23:49And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
23:55That's exactly right. They've done a 180 because of numbers like this.
23:58Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs because President Trump has argued that it will.
24:07Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
24:36the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year at this point, it was just a one-point lead for the Democrats. Look at where we are now.
25:02That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
25:22More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party chances are very very high, very much near 100% you're going to vote for that party.
25:33What about some of the groups that had been seen as crucial to President Trump?
25:37Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus 8 Democrat in early September of 2026?
25:49Well, it all comes down to this split non-ol college. Look, in 2024, Democrats had the college vote right on party ID.
25:57That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
26:2616-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
26:41They now lead by a point on this metric and that has allowed their lead to balloon to eight points overall when it comes to party identification.
26:49That's a big shift among a big group for President Trump.
26:52That's exactly right. A group that backed President Trump, big league has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
27:00All right. So, how much does party ID really matter if we look back in the midterms this century?
27:04Yeah. Okay. So, you see, right, two, one, and then the eight lead are the Democrats, right? Why does this matter?
27:11Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Let him party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
27:24They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the House. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
27:51for a midterm cycle this century is 18%.
27:54But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
28:01Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
28:12And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
28:20Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
28:37They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
28:53Every other election, every other election, midterm election this century at this point, what you're looking at, the prior month changed, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
29:05All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
29:10Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war. Overall, 71% say they are unwilling compared to just 29% who say willing. Look at this number among independents.
29:2681% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
29:35What are the prediction markets saying about where gas prices might be by election day?
29:38Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uhuh. Cowshi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
29:57These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
30:26polls across 2026. He's 42 points underwater. That is a switcheroo Kate Baldwin of what? 43 points shifting away from the president of United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
30:45What on what issues?
30:47Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which workingclass Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
31:15now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 plus 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
31:44of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree, he has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
32:08What are you where are you particularly watching to see how this plays out in the midterms? Right. Exactly. Right. So this is the president of the United States.
32:14You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
32:43Democrat James Telerico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Terico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
33:13Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
33:21In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest name see for at minimum discussions about regulating it. 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
33:37Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lorie Tran, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
34:06Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
34:34capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Amade now says he is going to start doing and wants the others to do as well.
34:56Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
35:26there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
35:37Yeah. I mean it's incredible gi given just last week, you know, you have people coming out saying entire all of human race is threatened. Yeah. and and and Amade comes out with this proposal on on Saturday that you're no members of Congress coming back until after the afterction the elections.
35:58That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled, uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
36:15uh you know the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China. This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate."
36:36What do you say to that argument?
36:38I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards. And I'm worried about
37:06entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
37:17Uh, Congressman, I really appreciate it.
37:19Thanks so much for your time.
37:21Thank you, Anderson.
37:22Lori Tran, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
37:30Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
37:39Anything's headed? Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
37:56Um but you know there has to it has to result in real legislation real global uh cooperation especially you know there's this upcoming visit by shei but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
38:17I want to play you something that um you've all Noah Harari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
38:27You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
38:54this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone who
39:23understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much we put in the hands of the people and their elected officials. And I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide something that's so
39:52important if it's as important as they say it's too important for them to run very much like nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oppenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
40:21Garrett Swisser, thanks so much.
40:25The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
40:53right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
41:11Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
41:16And why is that? Why is that a good idea?
41:18Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
41:34If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
41:47Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
42:09I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
42:28Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence uh
42:57a pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails so to speak.
43:06What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
43:16I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a model. AI is like a car and innovation
43:46is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
44:08That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
44:23So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
44:34Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily and they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
44:57They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing a from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time? You said something
45:27earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
45:57If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
46:27we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
46:47Would you say that you you're an optimist when it comes to these issues?
46:51Um, no. Um, I'm hopeful.
46:55But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
47:24stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
47:41So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
48:03Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
48:32been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the hugging face hack where open AI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
49:01that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet. And they still need human intervention to help them learn.
49:30But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
49:59are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent, who's going where, there's a small pool of people to choose from.
50:14These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI, as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
50:43is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
51:05President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
51:14The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's not approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result is low. How about among independents?
51:39You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
52:02Okay, we're going to go from the galaxy back to here on Earth.
52:06We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
52:18Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
52:34I mean, just look here. Okay, local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
53:03we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
53:27And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
53:33That's exactly right. They've done a 180 because of numbers like this.
53:36Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs because President Trump has argued that it will.
53:45Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
54:14the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year, at this point, it was just a one-point lead for the Democrats. Look at where we are now.
54:39That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
55:00More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party.
55:11What about some of the groups that had been seen as crucial to President Trump?
55:15Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus 8 Democrat in early September of 2026?
55:27While it all comes down to this split non-ol college look in 2024, Democrats had the college vote right on party ID.
55:35That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
56:0416-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
56:19They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
56:26That's a big shift among a big group for President Trump.
56:30That's exactly right. A group that backed President Trump, big league, has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
56:37All right. So, how much does party ID really matter if we look back in the midterms this century?
56:42Yeah. Okay. So, you see, right, two, one, and then the eight lead are the Democrats, right? Why does this matter?
56:49Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Let him party ID. They won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
57:02They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average for a midterm cycle this century is 18%.
57:32But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
57:39Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
57:50And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
57:57Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
58:15They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
58:31Every other election, every other election, midterm election this century at this point, what you're looking at the prior month changed, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
58:43All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
58:48Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war.
58:58Overall, 71% say they unwilling compared to just 29% who say willing. Look at this number among independents. 81% 81% four in five say they are unwilling to pay for higher gas prices because the war in Iran. President Trump's argument not winning the day.
59:12What are the prediction markets saying about where gas prices might be by election day?
59:16Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uhuh. Cowshi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
59:35These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
1:00:04polls across 2026. He's 42 points underwater. That is a switcheroo Kate Baldwin of what? 43 points shifting away from the president of United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
1:00:23What on what issues?
1:00:25Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which working-class Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
1:00:53now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 + 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
1:01:22of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree. He has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
1:01:45What are you where are you particularly watching to see how this plays out in the midterms? Right. Exactly. Right. So this is the president of the United States.
1:01:52You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
1:02:21Democrat James Telico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Terico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
1:02:50Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
1:02:58In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest names see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
1:03:15Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lori Trrahan, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
1:03:43Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
1:04:12capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Ammed now says he is going to start doing and wants the others to do as well.
1:04:34Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
1:05:03there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
1:05:14Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened and and and Ammed comes out with this proposal on on Saturday that you're no members of Congress coming back until after thection the elections.
1:05:35That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
1:05:52Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China.
1:06:06This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate." What do you say to that argument?
1:06:16I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards. And I'm worried about
1:06:43entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
1:06:55Uh, Congressman, I really appreciate it.
1:06:57Thanks so much for your time.
1:06:58Thank you, Anderson.
1:07:00Lori Trahan, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
1:07:08Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
1:07:17Anything's headed. Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
1:07:34Um but you know there has to it has to result in real legislation, real global uh cooperation, especially you know there's this upcoming visit by she but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
1:07:55I want to play you something that um you've all Noah Harrari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
1:08:05You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
1:08:31this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone one who
1:09:01understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials and I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
1:09:29something that's so important if it's as important as they say it's too important for them to run very much like a nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
1:09:59Garrett Swisser, thanks so much.
1:10:02The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
1:10:31right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
1:10:48Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
1:10:53And why is that? Why is that a good idea?
1:10:56Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the Hugging Face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
1:11:11If we're going to get these swarms of rogue AIs conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
1:11:24Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
1:11:47I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake corrupt elections. Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the
1:12:16AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence, a pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails, so to speak.
1:12:44What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
1:12:54I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
1:13:20model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
1:13:46That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
1:14:01So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
1:14:12Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily. And they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
1:14:35They're aligned. The interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing air from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time?
1:15:04You said something earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
1:15:35If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
1:16:05we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
1:16:24Would you say that you you're an optimist when it comes to these issues?
1:16:29Um, no. Um, I'm hopeful.
1:16:32But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
1:17:01stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
1:17:19So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
1:17:40Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
1:18:10been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the Hugging Face hack where OpenAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
1:18:39that these concerns keep them up at night that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
1:19:08But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
1:19:36are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent and who's going where. There's a small pool of people to choose from.
1:19:51These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI as AI is able to fix itself. You won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
1:20:21is for a lot of them this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done leave in a public manner and try to pressure things from the outside.
1:20:43President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
1:20:52The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's net approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result.
1:21:15How about among independents? You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You would have to combine Jacustoe with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
1:21:40Okay, we're going to go from the galaxy back to here on Earth.
1:21:44We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
1:21:56Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
1:22:12I mean, just look here. Okay? Local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
1:22:40we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer. The American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
1:23:05And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
1:23:10That's exactly right. They've done a 180 because of numbers like this.
1:23:13Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs? Because President Trump has argued that it will.
1:23:22Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
1:23:52the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024, in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year, at this point, it was just a one-point lead for the Democrats. Look at where we are now.
1:24:17That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
1:24:37More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party. What about some of the groups that had been seen as crucial to President Trump?
1:24:53Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus8 Democrat in early September of 2026?
1:25:05Well, it all comes down to this split non-ol college. Look, in 2024, Democrats had the college vote right on party ID.
1:25:12That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be whoa again a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
1:25:4116point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column. They now lead by a point on this metric and that has allowed their lead to balloon to eight points overall when it comes to party identification.
1:26:04That's a big shift among a big group for President Trump.
1:26:07That's exactly right. A group that backed President Trump big league has been backing Republicans. They are abandoning the Republican party going into the Democratic camp. All right. So, how much does party ID really matter if we look back at the midterms this century?
1:26:19Yeah. Okay. So, you see, right, two, one, and then the eight lead for the Democrats, right? Why does this matter?
1:26:26Why does this matter? Because in midterms this century, take a look here. The Dems win the House. 2000 midterms. Lenin party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
1:26:39They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of a midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
1:27:06for a midterm cycle this century is 18%.
1:27:09But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
1:27:17Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
1:27:28And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
1:27:35Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
1:27:52They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
1:28:08Every other election, every other election, midterm election this century at this point, what you're looking at the prior month change, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
1:28:21All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
1:28:25Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make, but the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war. Overall, 71% say they're unwilling compared to just 29% who say, "Well, look at this number among independents."
1:28:4281% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
1:28:50What are the prediction markets saying about where gas prices might be by election day?
1:28:53Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uh cowi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
1:29:13These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
1:29:42polls across 2026. He's 42 points underwater. That is a switcheroo ballwin of what? 43 points shifting away from the president of the United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
1:30:00What on what issues?
1:30:02Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue, but one issue on which working-class Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay? non-ol Latinos trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
1:30:30now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. Minus 47 plus 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
1:30:59of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree, he has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
1:31:23What are you where are you particularly watching to see how this plays out in the midterm?
1:31:27Right. Exactly right. So this is the president of the United States. You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a
1:31:57key Senate race going on between Democrat James Telerico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Telerico along with those massive gains among whites with a college degree in Texas
1:32:26and that is a winning formula for Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
1:32:36In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest name see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
1:32:53Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lorie Tran, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
1:33:21Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
1:33:50capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Amade now says he is going to start doing and wants the others to do as well.
1:34:11Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
1:34:41there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
1:34:52Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened. Yeah. And and and Amade comes out with this proposal on on Saturday that you're no members of Congress coming back until after the afterction the elections.
1:35:13That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
1:35:30Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China. This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate."
1:35:52What do you say to that argument?
1:35:53I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards and I'm worried about
1:36:21entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
1:36:33Uh, Congressman, I really appreciate it.
1:36:35Thanks so much for your time.
1:36:36Thank you, Anderson.
1:36:37Lori Tran, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
1:36:46Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
1:36:54Anything's headed? Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
1:37:12Um but you know there has to it has to result in real legislation real global uh cooperation especially you know there's this upcoming visit by shei but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
1:37:33I want to play you something that um you all know Harari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
1:37:43You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
1:38:09this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point I think crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and
1:38:38anyone who understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials. And I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
1:39:07something that's so important if it's as important as they say it's too important for them to run very much like nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
1:39:36Garrett Swooer, thanks so much.
1:39:40The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
1:40:09right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
1:40:26Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
1:40:31And why is that? Why is that a good idea?
1:40:34Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
1:40:49If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
1:41:02Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing, they throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
1:41:25I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
1:41:43Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence uh
1:42:12a pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails so to speak.
1:42:22What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
1:42:31I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a model that AI is like a car and
1:43:01innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
1:43:23That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
1:43:38So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
1:43:50Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily and they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
1:44:13They're aligned. The interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when the interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing AI from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time?
1:44:42You said something earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
1:45:13If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
1:45:42we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
1:46:02Would you say that you you're an optimist when it comes to these issues?
1:46:06Um, no. Um, I'm hopeful.
1:46:10But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
1:46:39stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
1:46:56So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
1:47:18Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
1:47:48been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the Hugging Face hack where OpenAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
1:48:16that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
1:48:46But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
1:49:14are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent and who's going where. There's a small pool of people to choose from.
1:49:29These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
1:49:59is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
1:50:20President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
1:50:29The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's not approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result is low. How about among independents?
1:50:54You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
1:51:18Okay, we're going to go from the galaxy back to here on Earth.
1:51:21We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
1:51:34Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
1:51:49I mean, just look here. Okay? Local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
1:52:18we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
1:52:42And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
1:52:48That's exactly right. They've done a 180 because of numbers like this.
1:52:51Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs because President Trump has argued that it will.
1:53:00Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
1:53:30the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year, at this point, it was just a onepoint lead for the Democrats. Look at where we are now.
1:53:55That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
1:54:15More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party.
1:54:26What about some of the groups that had been seen as crucial to President Trump?
1:54:30Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus 8 Democrat in early September of 2026?
1:54:42Well, it all comes down to this split non-ol college. Look, in 2024, Democrats had the college vote right on party ID.
1:54:50That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
1:55:1916-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
1:55:35They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
1:55:42That's a big shift among a big group for President Trump.
1:55:45That's exactly right. A group that backed President Trump, big league, has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
1:55:53All right. So, how much does party ID really matter if we look back in the midterms this century?
1:55:57Yeah. Okay. So, you see, right, two, one, and then the eight lead are the Democrats, right? Why does this matter?
1:56:04Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Let him party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
1:56:17They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
1:56:44for a midterm cycle this century is 18%.
1:56:47But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
1:56:54Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
1:57:05And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years? Yeah, normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
1:57:30They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact, the relief not there at all. In fact, there is more pain at the pump now.
1:57:46Every other election, every other election, midterm election this century at this point, what you're looking at, the prior month changed, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
1:57:59All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
1:58:03Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. Willing to pay more because of the Iran war. Overall, 71% say they are unwilling compared to just 29% who say willing. Look at this number among independents.
1:58:1981% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
1:58:28What are the prediction markets saying about where gas prices might be by election day?
1:58:31Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uhuh. Cowshi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
1:58:50These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
1:59:19polls across 2026. He's 42 points underwater. That is a switcheroo Kate Baldwin of what? 43 points shifting away from the president of United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
1:59:38What on what issues?
1:59:40Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which workingclass Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
2:00:08now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 plus 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
2:00:37of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree, he has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
2:01:01What are you where are you particularly watching to see how this plays out in the midterms? Right. Exactly. Right. So this is the president of the United States.
2:01:07You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
2:01:36Democrat James Telico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Terico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
2:02:06Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
2:02:14In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest names see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
2:02:30Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lori Trrahan, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
2:02:59Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh OpenAI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
2:03:28capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Ammed now says he is going to start doing and wants the others to do as well.
2:03:49Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that just that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
2:04:19there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
2:04:30Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened and and and Ammed comes out with this proposal on on Saturday that you're no members of Congress coming back until after thection the elections.
2:04:51That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
2:05:08Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China.
2:05:22This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate." What do you say to that argument?
2:05:31I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards. And I'm worried about
2:05:59entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
2:06:10Uh, Congressman, I really appreciate it.
2:06:12Thanks so much for your time.
2:06:14Thank you, Anderson.
2:06:15Lori Trahan, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
2:06:23Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
2:06:32Anything's headed. Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
2:06:50Um, but you know there has to it has to result in real legislation, real global uh cooperation, especially you know there's this upcoming visit by she, but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
2:07:10I want to play you something that um you've all Noah Harrari said in late July about AI which I yeah I think it's really I mean really fascinating. I just want to play this.
2:07:20You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
2:07:47this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point I crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone one who
2:08:16understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials and I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
2:08:45something that's so important if it's as important as they say it's too important for them to run very much like a nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oppenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
2:09:14Garrett Swisser, thanks so much.
2:09:18The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
2:09:46right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
2:10:04Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
2:10:09And why is that? Why is that a good idea?
2:10:12Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
2:10:27If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
2:10:40Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
2:11:03I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
2:11:21Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence uh
2:11:50a pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails so to speak.
2:12:00What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
2:12:09I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
2:12:36model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
2:13:01That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
2:13:16So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
2:13:28Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily and they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
2:13:51They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing AI from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time? You said something
2:14:20earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
2:14:51If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
2:15:20we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
2:15:40Would you say that you you're an optimist when it comes to these issues?
2:15:44Um, no. Um, I'm hopeful.
2:15:48But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
2:16:17stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Hadas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
2:16:34So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
2:16:56Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
2:17:26been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the hugging face hack where open AAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
2:17:54that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
2:18:24But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
2:18:52are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent, who's going where, there's a small pool of people to choose from.
2:19:07These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI, as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
2:19:36is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside. Becky, President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put
2:20:05not anywhere close together on this one.
2:20:07The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's not approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result is low. How about among independents?
2:20:32You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
2:20:55Okay, we're going to go from the galaxy back to here on Earth.
2:20:59We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
2:21:11Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
2:21:27I mean, just look here. Okay, local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
2:21:56we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
2:22:20And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
2:22:26That's exactly right. They've done a 180 because of numbers like this.
2:22:29Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs because President Trump has argued that it will.
2:22:38Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
2:23:07the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year at this point, it was just a one-point lead for the Democrats. Look at where we are now.
2:23:33That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
2:23:53More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party chances are very very high, very much near 100% you're going to vote for that party.
2:24:04What about some of the groups that had been seen as crucial to President Trump?
2:24:08Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus 8 Democrat in early September of 2026?
2:24:20Well, it all comes down to this split non-ol college. Look, in 2024, Democrats had the college vote right on party ID.
2:24:28That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
2:24:5716-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
2:25:12They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
2:25:20It's a big shift among a big group for President Trump.
2:25:23That's exactly right. A group that backed President Trump, big league has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
2:25:30All right. So, how much does party ID really matter if we look back in the midterms this century?
2:25:35Yeah. Okay. So, you see, right, two, one, and then the eight lead are the Democrats, right? Why does this matter?
2:25:42Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Let him party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
2:25:55They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of a midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
2:26:22for a midterm cycle this century is 18%.
2:26:25But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
2:26:32Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
2:26:43And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
2:26:50Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
2:27:08They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
2:27:24Every other election, every other election, midterm election this century at this point, what you're looking at the prior month changed, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
2:27:36All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
2:27:41Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war. Overall, 71% say they are unwilling compared to just 29% who say willing. Look at this number among independents.
2:27:5781% four in five say they are unwilling to pay for higher gas prices because the war in Iran. President Trump's argument not winning the day.
2:28:06What are the prediction markets saying about where gas prices might be by election day?
2:28:09Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uhuh. Cowshi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big league problems for Republicans running. These numbers hold. You can say adios amigos.
2:28:31Goodbye to that Republican majority in the House and maybe the Senate as well.
2:28:34Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS polls across 2026. He's 42 points underwater. That is a switcheroo Kate
2:29:03Baldwin of what? 43 points shifting away from the president of United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
2:29:16What on what issues?
2:29:18Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which workingclass Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
2:29:46now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 + 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
2:30:15of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree. He has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
2:30:38What are you where are you particularly watching to see how this plays out in the midterms? Right. Exactly. Right. So this is the president of the United States.
2:30:45You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know what? First of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
2:31:14Democrat James Telico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Terico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
2:31:44Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
2:31:51In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest names see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
2:32:08Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lori Trrahan, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
2:32:37Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
2:33:05capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Ammed now says he is going to start doing and wants the others to do as well.
2:33:27Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
2:33:56there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
2:34:07Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened and and and Ammed comes out with this proposal on on Saturday that you're no members of Congress coming back until after thection the elections.
2:34:29That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
2:34:46Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China.
2:35:00This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate." What do you say to that argument?
2:35:09I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards. And I'm worried about
2:35:36entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
2:35:48Uh, Congressman, I really appreciate it.
2:35:50Thanks so much for your time.
2:35:52Thank you, Anderson.
2:35:53Lori Trahan, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
2:36:01Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
2:36:10Anything's headed. Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
2:36:27Um, but you know, there has to it has to result in real legislation, real global uh cooperation, especially, you know, there's this upcoming visit by Shei, but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So, I mean, we have to it has to translate into real guard rails and and that's not going to happen till after the election.
2:36:48I want to play you something that um you've all Noah Harrari said in late July about AI, which I Yeah, I think it's really I mean, really fascinating. I just want to play this.
2:36:58You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
2:37:24this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point I crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone one who
2:37:54understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials and I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
2:38:22something that's so important if it's as important as they say it's too important for them to run very much like a nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oppenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
2:38:52Garrett Swisser, thanks so much.
2:38:56The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um we want to talk through some of these proposals uh before Congress
2:39:24right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
2:39:42Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
2:39:47And why is that? Why is that a good idea?
2:39:49Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
2:40:04If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
2:40:18Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
2:40:40I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
2:40:59Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence, a
2:41:28pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails, so to speak.
2:41:37What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
2:41:47I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
2:42:13model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
2:42:39That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
2:42:54So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
2:43:05Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily. And they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it?
2:43:28They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when the interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing air from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time? You said something
2:43:58earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
2:44:28If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
2:44:58we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
2:45:17Would you say that you you're an optimist when it comes to these issues?
2:45:22Um, no. Um, I'm hopeful.
2:45:26But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
2:45:54stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Hadas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
2:46:12So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
2:46:34Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
2:47:03been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the hugging face hack or open AI agents created. A swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
2:47:32that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
2:48:01But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
2:48:30are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent, who's going where, there's a small pool of people to choose from.
2:48:44These companies really value their talent and they want to keep them. and they want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
2:49:14is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
2:49:36President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
2:49:45The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's net approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result.
2:50:08How about among independents? You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
2:50:33Okay, we're going to go from the galaxy back to here on Earth.
2:50:37We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
2:50:49Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
2:51:05I mean, just look here. Okay? Local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
2:51:34we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
2:51:58And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
2:52:04That's exactly right. They've done a 180 because of numbers like this.
2:52:07Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs? Because President Trump has argued that it will.
2:52:16Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
2:52:45the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September.
2:53:02This foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year at this point, it was just a one-point lead for the Democrats. Look at where we are now.
2:53:10That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
2:53:31More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party.
2:53:42What about some of the groups that had been seen as crucial to President Trump?
2:53:46Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus8 Democrat in early September of 2026?
2:53:58While it all comes down to this split non-ol college look in 2024, Democrats had the college vote right on party ID.
2:54:06That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
2:54:3416-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
2:54:50They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
2:54:57It's a big shift among a big group for President Trump.
2:55:01That's exactly right. A group that backed President Trump, big league, has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
2:55:08All right. So, how much does party ID really matter if we look back in the midterms this century?
2:55:12Yeah. Okay. So, you see, right, two, one, and then the eight lead for the Democrats, right? Why does this matter?
2:55:20Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Lenin party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
2:55:33They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of a midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
2:56:00for a midterm cycle this century is 18%.
2:56:03But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
2:56:10Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jump.
2:56:21And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
2:56:28Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
2:56:45They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
2:57:02Every other election, every other election, midterm election this century at this point, what you're looking at the prior month change, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
2:57:14All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
2:57:18Yeah. Okay. That's an argument that President Trump has made. He liked those Republicans running for Congress to make, but the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war.
2:57:29Overall, 71% say they are unwilling compared to just 29% who say will look at this number among independents. 81% 81% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
2:57:43What are the prediction market saying about where gas prices might be by election day?
2:57:46Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uh cowi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
2:58:06These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos. Goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
2:58:35polls across 2026. He's 42 points underwater. That is a switcheroo kate Baldwin of what? 43 points shifting away from the president of the United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
2:58:54What on what issues?
2:58:55Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which working-class Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
2:59:24now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. Minus 47 plus 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
2:59:52of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree. He has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
3:00:16What are you where are you particularly watching to see how this plays out in the midterm?
3:00:20Right. Exactly right. So this is the president of the United States. You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a
3:00:50key Senate race going on between Democrat James Telerico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Telerico along with those massive gains among whites with a college degree in Texas
3:01:20and that is a winning formula for Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
3:01:29In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest name see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
3:01:46Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lorie Tran, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
3:02:14Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
3:02:43capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Amade now says he is going to start doing and wants the others to do as well.
3:03:04Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that's just that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
3:03:34there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
3:03:45Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened. Yeah. And and and Amade comes out with this proposal on on Saturday that you're no members of Congress coming back until after the afterction the elections.
3:04:06That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
3:04:23Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China. This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate."
3:04:45What do you say to that argument?
3:04:47I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards and I'm worried about
3:05:14entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
3:05:26Uh, Congressman, I really appreciate it.
3:05:28Thanks so much for your time.
3:05:29Thank you, Anderson.
3:05:31Lori Tran, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
3:05:39Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
3:05:48Anything's headed? Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
3:06:05Um but you know there has to it has to result in real legislation real global uh cooperation especially you know there's this upcoming visit by shei but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
3:06:26I want to play you something that um you all know Harari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
3:06:36You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
3:07:02this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and
3:07:31anyone who understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much we put in the hands of the people and their elected officials. And I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
3:08:00something that's so important if it's as important as they say it's too important for them to run very much like nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
3:08:30Garrett Swisser, thanks so much.
3:08:33The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
3:09:02right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
3:09:19Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
3:09:24And why is that? Why is that a good idea?
3:09:27Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
3:09:42If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
3:09:55Next is this AI kill switch bill. So that has companies with the ability, right? AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
3:10:18I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
3:10:37Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence uh
3:11:05a pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails so to speak.
3:11:15What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
3:11:25I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
3:11:51model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
3:12:17That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
3:12:32So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
3:12:43Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily and they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
3:13:06They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing a from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time? You said something
3:13:35earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
3:14:06If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
3:14:36we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
3:14:55Would you say that you you're an optimist when it comes to these issues?
3:15:00Um, no. Um, I'm hopeful.
3:15:03But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
3:15:32stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
3:15:49So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
3:16:11Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
3:16:41been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the hugging face hack where open AI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by human. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
3:17:10that these concerns keep them up at night, that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet. And they still need human intervention to help them learn.
3:17:39But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
3:18:07are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent, who's going where, there's a small pool of people to choose from.
3:18:22These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI, as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
3:18:52is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
3:19:14President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
3:19:22The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's not approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result is low. How about among independents?
3:19:47You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
3:20:11Okay, we're going to go from the galaxy back to here on Earth.
3:20:14We're back on planet Earth because look, at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
3:20:27Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
3:20:43I mean, just look here. Okay, local data centers prefer candidates for Congress oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? just 15% when I can easily do it like this get up to the support level 10 and then the additional five if I can do it on two hands you know it's low just 15% but again you think that 15% is low how about we come over to this side of the screen with independence now
3:21:11we're just talking about 8% prefer candidates who will in fact support the establishment of data centers the eight you can literally do it on two hands the opposition look at that among independents through the roof 62% it is skyhigh in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
3:21:36And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
3:21:41That's exactly right. They've done a 180 because of numbers like this.
3:21:44Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs because President Trump has argued that it will.
3:21:53Yeah. President Trump has said that the establishment of data centers, AI, will in fact increase the number of jobs. The American people, no, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
3:22:23the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024 in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year at this point, it was just a one-point lead for the Democrats. Look at where we are now.
3:22:48That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
3:23:08More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party.
3:23:20What about some of the groups that had been seen as crucial to President Trump?
3:23:24Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus 8 Democrat in early September of 2026?
3:23:36While it all comes down to this split non-OL college look in 2024, Democrats had the college vote right on party ID.
3:23:43That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
3:24:1216-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
3:24:28They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
3:24:35That's a big shift among a big group for President Trump.
3:24:38That's exactly right. A group that backed President Trump, big league, has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
3:24:46All right. So, how much does party ID really matter if we look back in the midterms this century?
3:24:50Yeah. Okay. So, you see, right, two, one, and then the eight lead are the Democrats, right? Why does this matter?
3:24:57Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Let him party ID. They won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
3:25:10They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of a midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average for a midterm cycle this century is 18%.
3:25:40But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
3:25:48Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jumps.
3:25:58And this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
3:26:06Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
3:26:23They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
3:26:39Every other election, every other election, midterm election this century at this point, what you're looking at the prior month changed, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
3:26:52All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
3:26:56Yeah. Okay. That's an argument that President Trump has made. He like those Republicans running for Congress to make. But the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war. Overall, 71% say they are unwilling compared to just 29% who say willing. Look at this number among independents.
3:27:1281% four in five say they are unwilling to pay for higher gas prices because the war in Iran. President Trump's argument not winning the day.
3:27:21What are the prediction markets saying about where gas prices might be by election day?
3:27:24Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uhuh. Cowshi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big league problems for Republicans running.
3:27:44These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
3:28:13polls across 2026. He's 42 points underwater. That is a switcheroo Kate Baldwin of what? 43 points shifting away from the president of United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
3:28:31What on what issues?
3:28:33Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which working-class Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump in immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
3:29:01now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 + 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
3:29:30of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree. He has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
3:29:54What are you where are you particularly watching to see how this plays out in the midterms? Right. Exactly. Right. So this is the president of the United States.
3:30:00You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
3:30:29Democrat James Telico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Terico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
3:30:59Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
3:31:07In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest names see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
3:31:24Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lori Trrahan, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
3:31:52Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
3:32:21capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Ammed now says he is going to start doing and wants the others to do as well.
3:32:42Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
3:33:12there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
3:33:23Yeah. I mean it's incredible gi given just the last week, you know, you have people coming out saying entire all of human race is threatened and and and Ammed comes out with this proposal on on Saturday that you're no members of Congress coming back until after thection the elections.
3:33:44That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
3:34:01Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China.
3:34:15This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate." What do you say to that argument?
3:34:24I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards. And I'm worried about
3:34:52entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
3:35:04Uh, Congressman, I really appreciate it.
3:35:06Thanks so much for your time.
3:35:07Thank you, Anderson.
3:35:08Lori Tran, thank you. Joining us now is CNN contributor and tech journalist Cara Swisser.
3:35:16Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
3:35:25Anything's headed. Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
3:35:43Um, but you know there has to it has to result in real legislation, real global uh cooperation, especially you know there's this upcoming visit by Shei, but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
3:36:04I want to play you something that um you've all Noah Harrari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
3:36:13You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
3:36:40this idea that yeah these companies are creating kind of AI superhuman super beings or gods and expecting them to cowtow to us seems I mean he makes a good point crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone one who
3:37:09understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials and I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide
3:37:38something that's so important if it's as important as they say it's too important for them to run very much like a nuclear a nuclear bomb or nuclear fision um and as you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
3:38:07Garrett Swisser, thanks so much.
3:38:11The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
3:38:40right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
3:38:57Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
3:39:02And why is that? Why is that a good idea?
3:39:05Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the Hugging Face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
3:39:20If we're going to get these swarms of rogue AIs conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
3:39:33Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing. They throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
3:39:56I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
3:40:14Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence, a
3:40:43pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails, so to speak.
3:40:53What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
3:41:02I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid to develop that before you know whether it's going to turn out nicely. Um, for that reason, we should certainly slow down the development of super intelligence. Slowing down the rest of AI is going to be difficult. The AI companies would like you to have a
3:41:29model. AI is like a car and innovation is like the accelerator. AI development is like the accelerator and regulation is like the brakes. So they want to sort of phrase regulation as a bad thing that slows everybody down. A much better model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
3:41:54That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China?
3:42:09So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things.
3:42:21Both the Chinese Communist Party and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily. And they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
3:42:44They're aligned. The interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned. On things like fake videos for corrupting elections, the interests are anti-aligned because all the countries are doing it to each other. But on things like preventing AI from taking over, no country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it. America doesn't want it. And so everybody can agree on that. The question is, can they do something in time? You said something
3:43:13earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it already is, right? It's helping to develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward
3:43:44If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in control. The mother's in charge, but the baby can control the mother by crying, for example. the mother just can't bear the baby crying. We need to develop since we're building this super intelligence,
3:44:13we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not sure we can do that.
3:44:33Would you say that you you're an optimist when it comes to these issues?
3:44:37Um, no. Um, I'm hopeful.
3:44:41But I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep alignment and safety and monitoring way ahead of capabilities to slow down or
3:45:10stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
3:45:27So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
3:45:49Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers, to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
3:46:19been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the Hugging Face hack where OpenAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by humans. That was the AI's interpretation of what they needed to do to ace a cyber security exam. AI scientists told me
3:46:47that these concerns keep them up at night that they are so worried about this and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym. Right now, an AI model isn't quite there yet, and they still need human intervention to help them learn.
3:47:17But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have any sort of control over it. And the other thing of why that matters, recursive self-improvement, because right now these AI scientists who are very outspoken, most industries, you wouldn't dare to be posting things on social media like all of these AI scientists
3:47:45are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists, they are getting paid a lot of money. There's a huge back and forth of, you know, AI talent and who's going where. There's a small pool of people to choose from.
3:48:00These companies really value their talent and they want to keep them. They want to keep them happy. And so right now they have leverage. They are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments
3:48:30is for a lot of them, this is not about money. This is not about their own personal wealth. They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
3:48:51President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one.
3:49:00The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here. Trump's net approval rating of AI. That's approval minus disapproval. Look at this. Overall, he's 34 points underwater. My goodness gracious. And somehow that's become worse over the course of the year because back in March he was closer around 25 points underwater. But you think this 34 points underwater result.
3:49:23How about among independents? You ain't seen nothing yet. Sarah's off on the side shaking her head. 61 points underwater. You'd have to combine Jacusto with the Little Mermaid to get as far underwater as Donald Trump is with independence on AI. As I said, the American people and independents especially absolutely despise President Trump when it comes to this issue that is becoming even more key as we lead up to the midterm elections.
3:49:49Okay, we're going to go from the galaxy back to here on Earth.
3:49:52We're back on planet Earth because look, the at one point on social media, Trump chided voters to support data centers. So, how are voters feeling about data centers in general? And did that make any difference?
3:50:04Yeah. Yeah. Yeah. Yeah. You know, you see these low numbers for Trump on AI. I'm going to highlight it right there. There we go. Low numbers on AI. Therefore, it shouldn't be so much of a surprise that the American people do not want those folks who are running for Congress to support the establishment of data centers in their local communities.
3:50:20I mean, just look here. Okay? Local data centers prefer candidates for Congress, oppose or support. Look at this opposition. 54% say they want those candidates running for Congress to oppose data centers. How about the support? Just 15%. When I can easily do it like this, get up to the support level 10 and then the additional five.
3:50:40If I can do it on two hands, you know, it's low. Just 15%. But again, you think that 15% is low. How about we come over to this side of the screen with independence. Now we're just talking about 8% prefer candidates who will in fact support the establishment of data centers. The eight, you can literally do it on two hands. The opposition, look at that, among independents through the roof. 62%.
3:51:02It is skyhigh. in terms of what they prefer, the American people prefer their candidates doing running for local office, running for Congress when it comes to data centers. They want to oppose oppose oppose.
3:51:13And it's no surprise that we've seen some Republicans come out very strongly and say, "Hey, we've got to rethink all of this."
3:51:19That's exactly right. They've done a 180 because of numbers like this.
3:51:22Yeah. Do do what are voters thinking about sort of whether AI is going to increase the number of jobs? Because President Trump has argued that it will.
3:51:31Yeah. President Trump has said that the establishment of data centers AI will in fact increase the number of jobs the American people. No, no, they do not buy it. Uh-uh. This is what they say. Just take a look here. What will AI do to jobs in America? Reduce 64% increase just 19%. But again, look at this side of the screen. Among independents, 72% of independents say that in fact AI will reduce jobs in America compared to just what is it 10. And then the additional four say increase. President Trump on
3:52:00the wrong side of this issue. I'm a Republican running for Congress this fall. I am running away from President Trump when it comes on AI and data centers because simply put, he's on the wrong side of the issue as far as voters are concerned. What you see here is a fundamental shift going on here. You know, back in 2024, in early September, this foreshadowed Donald Trump's victory. Republicans led on party ID by two points. Last year, at this point, it was just a one-point lead for the Democrats. Look at where we are now.
3:52:26That Democratic advantage up like a rocket. These are very good numbers for Democrats. Now up by eight points in terms of party ID. We are not just talking about the top line here in terms of the generic ballot moving. We are talking about fundamentally underneath the fundamentals are shifting in the Democrats direction on this very key metric.
3:52:46More people are saying they're Democrat Democrats than they are Republicans. And of course in this day and age when you say you belong to one party, chances are very very high, very much near 100% you're going to vote for that party.
3:52:57What about some of the groups that had been seen as crucial to President Trump?
3:53:01Yeah, exactly right. Why Why has this number shifted from plus two Republican in early September 2024 to now plus8 Democrat in early September of 2026?
3:53:13Well, it all comes down to this split non-ol college. Look, in 2024, Democrats had the college vote right on party ID.
3:53:21That is those who had graduated college were 14 points more likely to identify as Democrats than Republicans. But the non-ol vote is a larger portion of the electorate and they were 13 po points more likely to identify as Republicans than Democrats. Come over to here. What has happened? Johnny be. Whoa. Again, a big league shift. We are talking about a 14point shift in the Democrats direction. Now they lead on party ID among those who did not graduate college by a point. You combine that with this
3:53:5016-point advantage. It really didn't move among college. The reason why Democrats are doing so well on the fundamental question of whether or not you identify as a Democrat or a Republican at this point, at least according to the Times, is because those who didn't graduate college have shifted tremendously into the Democratic column.
3:54:06They now lead by a point on this metric, and that has allowed their lead to balloon to eight points overall when it comes to party identification.
3:54:13That's a big shift among a big group for President Trump.
3:54:16That's exactly right. A group that backed President Trump, big league, has been backing Republicans. They are abandoning the Republican party going into the Democratic camp.
3:54:24All right. So, how much does party ID really matter if we look back at the midterms this century?
3:54:28Yeah. Okay. So, you see, right, two, one, and then the eight lead for the Democrats, right? Why does this matter?
3:54:35Why does this matter? Because in midterms this century, take a look here. The Dems win the House 20,000 midterms. Lenin party ID, they won the House in the two of the two times that that happened. When they didn't lead, that is they were tied or trailing Republicans.
3:54:48They never won the House. So what we see here is because Democrats leading on party ID that says that they're going to win the house. We are talking about record high spikes in gas prices. I mean just look at this. Okay, you know normally gas prices do go up during the course of an election year during the course of a midterm year but never this high. This is record-breaking. We are talking record-breaking for a midterm this century. Changing gas prices from January 1 to September 14th. The average
3:55:15for a midterm cycle this century is 18%.
3:55:18But look at this on the right side of your screen. 52% rise this year. 52% rise. I said up like a rocket, right?
3:55:25Look at that. My goodness gracious. By far the highest on record. The prior high was just 27%. This is nearly double the prior record high. We are talking really really large jump and this is for the year this election season. If you're thinking about heading past Labor Day into November, where does that compare to past years?
3:55:44Yeah. Normally what you see is in the prior month that is from August 14th to September 14th normally you go down. So again looking at 21st century midterm elections went down all of them prior to 2026. This year however the gas prices have actually been going up up and away.
3:56:01They have actually risen up 6% over the last month. So normally you're getting hey there's some some relief going on with gas gas prices as we go into the midterm election. Not this year. In fact the relief not there at all. In fact, there is more pain at the pump now.
3:56:17Every other election, every other election, midterm election this century at this point, what you're looking at the prior month change, the gas prices all went down. This year, we're seeing the exact opposite. Gas prices are going up.
3:56:30All right. Now, what do people say about, oh, you know, this is the war in Iran. Maybe it's worth it.
3:56:34Yeah. Okay. That's an argument that President Trump has made. He liked those Republicans running for Congress to make, but the American people, uh-uh, they don't buy it. Look at this. willing to pay more because of the Iran war. Overall, 71% say they are unwilling compared to just 29% who say, "Well, look at this number among independents."
3:56:5081% four in five say they are unwilling to pay for higher gas prices because the war in Iran President Trump's argument not winning the day.
3:56:59What are the prediction markets saying about where gas prices might be by election day?
3:57:02Yeah. Okay. You might say, "Hey, maybe there'll be some relief. No more $4 gas prices." Uh uh cowi prediction market. Look at this chance gas prices on election day above $4 a gallon. Now a 59% chance compared to just a 30% chance at the beginning of the month. That is a doubling. This is big big big big big leap problems for Republicans running.
3:57:21These numbers hold. You can say adios amigos. Goodbye to that Republican majority in the House and maybe the Senate as well. Workingclass Latinos. He had a massive breakthrough in 2024. He won workingclass Latinos. He won non-ol Latino voters by a point over Kla Harris. That's a major breakthrough for a Republican president or Republican running for president. But that win that he had, adios, amigos, goodbye. See you later. Look at his net approval rating now at an average of polls, CNN, SSRS
3:57:50polls across 2026. He's 42 points underwater. That is a switcheroo kate Baldwin of what? 43 points shifting away from the president of the United States. his support rapidly declining, plummeting among working-class Latinos, a group that helped get him elected back in 2024.
3:58:09What on what issues?
3:58:11Yeah. Okay. So, you know, obviously we've spoken about the economy being an issue. But one issue on which working-class Latinos especially have turned against the president of the United States is the issue of immigration. There has been a massive, massive backlash, and you can see it here. Okay. non-ol Latinos Trump and immigration back in 2024 versus Kla Harris. He was more trusted on immigration amongst this group. Look at this. He won it by nine points. He was more trusted in this issue by nine points. Look at his net approval rating
3:58:39now on immigration when it comes to working-class Latinos. Those Latinos who didn't graduate college. It's 47 points underwater. Absolutely diving down in there into the Gulf of Mexico. Kate, you do the quick math with me here. - 47 + 9. That is a shift away from the president of the United States of 56 points. You don't see shifts away from a president of the United States of 56 points unless you have done something absolutely terribly wrong in the minds
3:59:08of the voters. And when it comes to working-class Latinos, that is Latinos without a college degree, he has done something horribly, horribly wrong. As I said at the top, no bueno for the president of the United States. He ran on immigration. He was more trusted in immigration than Kla Harris was with working-class Latinos. And now he is horribly underwater. We're talking 47 points underwater. These are terrible, terrible numbers.
3:59:32What are you where are you particularly watching to see how this plays out in the midterm? Right. Exactly right. So this is the president of the United States.
3:59:38You might say, "Hey, I'm a Republican running for Congress. I don't necessarily have to worry about the president of the United States. He's not on the ballot." But you know, first of all, the president wants to be on the ballot. And second off, you are running under the same party banner as the president of the United States. And we are seeing a massive shift away from Republican candidates. I mean, just take a look at Texas, okay? Texas, where, of course, Latinos make up a massive share of the population. Non-ol Latino voters in Texas. Look at this. Donald Trump won that vote by 15 points. But we have a key Senate race going on between
4:00:07Democrat James Telerico and Republican Ken Paxton. Look at this. James Telerico is up by 13 points among non-ol Latinos in the great state of Texas. That is a shift away of 28 points. And this is the type of movement that could help James Terico become the first Democrat to win a Senate race in Texas since 1988. You put together these big gains with working-class Latinos for James Telerico along with those massive gains among whites with a college degree in Texas and that is a winning formula for
4:00:37Democrats in Texas. Republicans should be running scared. Ken Paxton should be running scared. These are the types of numbers that make history for Democrats.
4:00:45In the middle of a week dominated by warnings about the potential dangers of Frontier AI and the need that some of its biggest name see for at minimum discussions about regulating it, 435 voices will be missing from that debate for the next 7 weeks. Those are of course the 435 members of Congress.
4:01:01Today, House Speaker Johnson sent them all home until after the midterms. Even though according to one member, there are dozens of pieces of proposed legislation now pending right now. Not to mention huge concerns about AI eliminating the human race. Joining us now is Massachusetts Democratic Congresswoman Lorie Tran, who is co-sponsoring a bipartisan measure on AI safety. Congresswoman, thanks for being here. So, why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?
4:01:30Uh, well, because first of all, right now we're only working on the honor system. Uh the only reason why we know there was a model that broke out of uh Open AI's testing environment is because they told us and that does not make anyone feel safe. Uh so we put pen to paper uh months ago on just basic guidelines and provisions on frontier AI safety. Frontier is these are the largest companies uh whose models are
4:01:58capable of catastrophic risk. uh mandatory transparency in incident reporting. So if a model does break out of the lab, they have to tell the government um independent evaluators that are embedded in the labs and checking their work and then which is something that Dario Amade now says he is going to start doing and wants the others to do as well.
4:02:20Absolutely. I mean the ground has completely shifted on AI safety. Um in the last few weeks we've seen safety researchers resign. We've seen models escape their labs and now we're seeing the three CEOs of these major uh frontier labs try to slow down. Um that should that's just that's a calling for Congress to get off the sidelines. Uh we don't have any legislation on these frontier labs. We have nothing in place right now to even have a kill switch if
4:02:50there is a model that is capable of a catastrophe. And so the fact that Speaker Johnson sent us home today is completely irresponsible. Uh there are so many members who want to act on this issue.
4:03:01Yeah. I mean it's incredible gi given just last week, you know, you have people coming out saying entire all of human race is threatened. Yeah. And and and Amade comes out with this proposal on on Saturday that you're no members of Congress coming back until after the afterction the elections.
4:03:22That's right. And uh and people are extremely uncomfortable. Uh you know, there have been letters that have been sent to the speaker to keep us in session, restore the two weeks of session that were cancelled. Uh you know, have a special session or even have members who have put pen to paper on legislation stay and work out a deal.
4:03:39Uh you know, the Frontier Act is a perfect starting point because it is bipartisan. It is something that uh could get the votes in both the House and the Senate. And well, let me uh let me ask you, what do you say to those who say, "Look, you know, we're in competition with China. This is an arms race. You can't trust uh you know, the Communist Party in China to to self-regulate."
4:04:00What do you say to that argument?
4:04:02I've been listening to that argument all week that if we slow down, China is going to race past us. Uh history tells a very different story. Uh the United States has always set the tough standards, the high standards on the most powerful technologies, nuclear weapons, biological and chemical weapons, aviation safety. Uh and frankly, safety and innovation are not in competition with each other. Uh that is exactly how we lead the world and we set the standards and I'm worried about
4:04:30entering into negotiations with China and others without having done the hard work back home. I mean, what kind of leverage do we have in those conversations if we're not bringing that to the table?
4:04:41Uh, Congressman, I really appreciate it.
4:04:43Thanks so much for your time.
4:04:45Thank you, Anderson.
4:04:46Lori Tran, thank you. Joining us now is CNN contributor and tech journalist, Cara Swisser, Cara, a lot has happened in just the last week regards to AI. Where do you think things stand right now in terms of Yeah. where this is heading?
4:05:03Anything's headed? Well, it's called hurry up and wait. You know, you sort of have to figure out what to do. Everyone sort of pushed the panic button and why now? Even though, as you know, you and I have talked about these problems for a long time. Um, you know, but these people are speaking up and you have to wonder why they're doing it and why at this moment. I have a lot of theories.
4:05:21Um but you know there has to it has to result in real legislation real global uh cooperation especially you know there's this upcoming visit by shei but I don't expect Donald Trump to do much of cooperation here um given what he said publicly. So I mean we have to it has to translate into real guard rails and and that's not going to happen till after the election.
4:05:41I want to play you something that um you all know Harari said in late July about AI which I yeah I think is really I mean really fascinating. I just want to play this.
4:05:51You listen to the to the narrative coming out of places like Silicon Valley and there is an inherent contradiction in the in the narrative. They tell us we will create a god and it will be our slave. And this doesn't make sense. If it is a god, it cannot remain a slave. And if it is a slave, it means it has no god-like abilities.
4:06:18this idea that yeah these companies are creating kind of AI superhum super beings or gods and expecting them to cowtow to us seems I mean he makes a good point crazy yeah I think you've all amazing you know he was very much like a craze in Silicon Valley over sapiens and then he wrote a second book where he sort of decrieded a lot of what was happening very early uh I think he's amazing historian and he understands history and anyone who
4:06:47understands history knows what they're saying is nonsense. This has happened before. Um the question is how much do we put in the hands of these companies and how much do we put in the hands of the people and their elected officials.
4:06:57And I think the second is where it needs to go. Um a lot of the rhetoric right now is we'll work on it and now we know and I don't trust them at all and neither does anyone else after the social media debacle that's happened over the last decade. And so the question is who's involved? who gets to decide and and do they get to decide something that's so important if it's as important as they say it's too important for them to run very much like nuclear a nuclear bomb or nuclear fision um and as
4:07:27you know as Scott and I have pointed out Oenheimer didn't sell his bombs to companies or companies didn't make the bombs and then decide whether where and when to deploy them um you don't want to nationalize this stuff but if it's as important as they say it is there has to be significant government involvement in how this stuff is rolled out.
4:07:45Garrett Swisser, thanks so much.
4:07:49The man considered to be the godfather of AI, Jeffrey Hinton. His research paved the way for how today's AI systems function. Jeffrey, thank you so much for being with us. Let's talk about uh these efforts by lawmakers to pass some form of AI regulation. There have been more than a hundred bills introduced in Congress to regulate AI over the past two years. None of them have actually passed. Um, we want to talk through some of these proposals uh before Congress
4:08:17right now and get your reaction as to how effective they might be. First, this uh bill that among other things requires top AI companies to allow independent verification organizations into their companies to make sure that their models are being developed safely. Do you see that as potentially effective?
4:08:35Yes, I think that's a very good idea. It's not the solution to everything, but it's a start.
4:08:40And why is that? Why is that a good idea?
4:08:42Well, at present, we've relied essentially on whistleblowers saying how things went rogue. And if you look at the hugging face incident, um it was reported by people in the companies, um some of whom left. Um but we need much more monitoring of that kind of thing.
4:08:58If we're going to get these swarms of rogue AIS conspiring with each other to break into other companies without the organizations knowing it's happening, um, we need people inside those companies who can say what's going on.
4:09:11Next is this AI kill switch bill. So that has companies with the ability, right, AI developers, they can they can do the killing, they throttle or they can suspend, they can shut down their AI. Then you have though DHS authorized to basically make the call on this. So they do have the control in that. What do you think about that possibility?
4:09:33I don't think it'll work in the long run. So there's a whole bunch of different risks of AI. There's the risks that come from bad actors using AI to do bad things like cause mass unemployment or create nasty viruses or do nasty cyber attacks or make videos for fake elections.
4:09:52Um, but then there's a different kind of risk which is AI itself getting much more intelligent and taking over AI taking control. A kill switch is no good for that because the AI will be much better than people at persuading people of things. It's already comparable with people at persuading you of things. When it's super intelligent, it'll be much better than people. So, it will be able to persuade the people in charge of the switch not to pull the switch. So to that point, Senator Bernie Sanders proposed a ban on super intelligence, a
4:10:21pause as well on uh the deployment of advanced AI models until regulators have developed more guard rails, so to speak.
4:10:30What are your thought on the idea of pumping the brakes on super intelligence? Is that something that can still be done at this point or is the genie out of the bottle?
4:10:40I think it's a good idea. For example, for super intelligence, we have no idea how we can stay in control or how we can get it even when it's in control to like us still and be nice to us. Um, so it seems very stupid model is um development of AI is like the accelerator and regulation is like the steering wheel. They want us to develop a very fast car with no steering wheel and that's a bad idea.
4:11:06That we can see how that would definitely be a bad idea. So is this whole conversation about regulating in America is it pointless without a global initiative without an agreement with China? So we will get I say we I used to think of North America as one thing but of course I'm from Canada and I no longer think of it as one thing. Um we will get agreement with China on some things. Both the Chinese Communist Party
4:11:34and the countries in North America um don't want terrorists being able to create nasty viruses easily. They don't want terrorists being able to do cyber attacks easily. And they particularly don't want AI taking over. So on that issue of should we slow down the development of super intelligent AI until we know how we can control it.
4:11:55They're aligned. Their interests are aligned. So parties will collaborate when their interests are aligned and they won't collaborate when their interests are anti-aligned on things like fake video. Country wants that. The Chinese Communist Party doesn't want it. Um the democracies don't want it.
4:12:09America doesn't want it. And so everybody can agree on that. The question is can they do something in time? You said something earlier that was fascinating and raises all sorts of philosophical questions as well about super intelligence and getting it to like us. I want you to to expand on that because at some point we're going to reach uh a a a point of no return where this thing is just going to sort of build itself. In some ways it it it already is right. It's helping to
4:12:38develop its own future. So what do you mean by super intelligence liking us? Is that possible? Are we headed toward Terminator? If you look around and ask what examples do we have of a more intelligent thing being controlled by a less intelligent thing, there aren't many out there, right? We don't notice that much because we're the apex intelligence on this planet. Um, but there is one example which is a baby and a mother. In a sense, they're both in
4:13:07control. The mother's in charge, but the baby can control the mother by crying, for example. The mother just can't bear the baby crying. We need to develop since we're building this super intelligence, we're designing it, we need to develop it in such a way that it likes us more than it likes itself and in such a way that even though it's much smarter than us, it wants us to realize our full potential. That's my belief. I think that's a better approach than thinking how do we always keep this smart thing under our thumb. I'm not
4:13:35sure we can do that.
4:13:37Would you say that you you're an optimist when it comes to these issues?
4:13:41Um, no. Um, I'm hopeful. Yeah, but I think we're at a very delicate point in history where what we do now is going to determine our future. And one thing we should be doing now is putting a lot of resources into figuring out how we can coexist with super intelligent AI. I am very confident in our company's ability, our industry's ability to do this safely to make sure that we keep
4:14:09alignment and safety and monitoring way ahead of capabilities to slow down or stop if we get to a point where we can't. Um, but I'm disappointed by how it's been framed. Well, CNN AI correspondent Adas Gold is in Montreal in Canada where a major AI conference is uh getting underway today.
4:14:31So, thank you for your time, Adas. Uh you have new reporting on how AI engineers and those at the forefront of creating this and let's call it super powerful technology uh in the country are feeling about the impact as this conversation about AI regulation, AI safety continues. What are you hearing from them?
4:14:53Yeah, so I spent the past few days reaching out to a bunch of contacts across the AI industry, mostly the people who are actually building this technology, the researchers, the engineers to ask them about this recent sort of floodgates that have been opened by Jacob Coxin, the former anthropic researcher who resigned, and these concerns about where AI is going and how it could hurt humanity. And what I found, which is something that I think a lot of us in the industry knew, is that these concerns are nothing new. And a lot of engineers and researchers had
4:15:22been concerned about AI's capabilities for some time. But things changed in the last few months, especially in the wake of incidents like the Hugging Face hack where OpenAI agents created a swarm of thousand more than a thousand agents communicated with each other, coordinated with each other to hack into a completely different company's servers, none of which was directed by humans. That was the AI's interpretation of what they needed to do to ace a cyber security. They are so worried about this
4:15:50and that this is a constant topic of conversation in the AI world. One thing that they are particularly concerned about is what's called recursive self-improvement. That is where the AI models will be capable of improving themselves. Just like her, you or I can, you know, go study and make ourselves better at something and go to the gym.
4:16:11Right now, an AI model isn't quite there yet. and they still need human intervention to help them learn. But these engineers say that it's only a matter of time before these AIs will be able to make themselves better, smarter, faster. And the concern is that it will get to the point where it'll be so fast and so good that humans just aren't able to monitor and intervene and have improvement. Because right now these AI scientists who are very outspoken most industries you wouldn't dare to be posting things on social media like all
4:16:38of these AI scientists are doing. They have leverage because of the talent war in the AI industry. The AI researchers and scientists they are getting paid a lot of money. There's a huge back and forth of you know AI talent and who's going where there's a small pool of people to choose from. These companies really value their talent and they want to keep them. They want to keep them happy. So right now they have leverage.
4:17:00they are putting pressure on their executives to do something, you know, not only about safety, but to speak out on what they're doing to keep things as safe as possible and to build in some guards. And the concern is that AI, as AI is able to fix itself, you won't need as many of these AI scientists and engineers and researchers and they will lose their leverage. And one thing that really also stood out to me about speaking to these engineers and and hearing their comments is for a lot of them, this is not about money. This is not about their own personal wealth.
4:17:30They are in this for real moral and ethical reasons. And some of them are really struggling with should they stay inside these companies and try to fix it from the inside or should they do what Jacob Coxin did and what some others have done, leave in a public manner and try to pressure things from the outside.
4:17:45Vicki, President Trump is in one galaxy and the American people are in a completely other galaxy. They're just simply put not anywhere close together on this one. The American people despise President Trump when it comes to his handling of artificial intelligence. You could see it right here.
