Return to Transcripts main page

CNN News Central

Trump Administration Files Final Reply Brief as Supreme Court Reviews Mail-In Ballot Order; Anthropic Says It Blocked Possible Attempts to Build Bioweapons; Trump Says Iran War Will End After Midterms and Gas Prices Will Fall; At Least Five Killed, Dozens Missing After Philippine Ferry Fire; White House Pulls "Build The Wall" Website Game After Warning From Tetris. Aired 1:30-2p ET

Aired September 10, 2026 - 13:30   ET

THIS IS A RUSH TRANSCRIPT. THIS COPY MAY NOT BE IN ITS FINAL FORM AND MAY BE UPDATED.

[13:30:00]

JOAN BISKUPIC, CNN CHIEF SUPREME COURT ANALYST: -- military. And. Right today, Boris, the Trump administration gave its final legal arguments saying, look, this is just a modest thing we're trying to do here. We're not trying to have the post office take over elections. We're just trying to have -- we're trying to protect against fraud.

There is no widespread fraud in mail-in ballots. But that's their argument.

(CROSSTALK)

BISKUPIC: And I think it's important for the Supreme Court to rule quickly because there's just such confusion out there.

BORIS SANCHEZ, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": Yeah, in some places mail-in ballots have already gone out.

BISKUPIC: That's right.

SANCHEZ: Joan Biskupic, thank you so much for the reporting.

BISKUPIC: Thank you.

SANCHEZ: Coming up, Anthropic says it has discovered another hacking incident involving one of its models and the latest breach comes after a former researcher quit the company over A.I. safety concerns. You'll hear from that researcher when we come back.

(COMMERCIAL BREAK)

SANCHEZ: Some Breaking News now. Anthropic, the maker of Claude, says that it blocked several potential plots this year by scientists who were using its A.I. models to conduct research that could have led to the creation of bioweapons.

BRIANNA KEILAR, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": This revelation, of course, coming just days after an Anthropic employee quit, warning that A.I. could kill all humans within the next 10 years.

[13:35:00]

When CNN asked that now former employee about his apocalyptic prediction, he pointed to one scenario being biological misuse.

(BEGIN VIDEO CLIP)

ANDERSON COOPER, CNN HOST OF "ANDERSON COOPER 360": How would that happen? Like what would that actually look like?

JACOB COXON, QUIT ANTHROPIC OVER A.I. FEARS: Yeah. So the precise scenario sounds a little bit like science fiction. It sounds like something that's not real, but I think it is frighteningly real and it's easiest to understand it if you think about real things that happened two months ago.

So two months ago, OpenAI agents, AIs, hacked into third-party infrastructure entirely at their own accord and this was like a concentrated hacking spree that they carried out of their own volition. And I think that if you extrapolate into the future, the level of capabilities of these AIs with the same independent volition, they could cause extreme havoc. For example, hacking critical infrastructure, building extinction-level bioweapons. There's a lot of ways that the A.I. could actuate itself in the world.

(END VIDEO CLIP)

KEILAR: Let's talk about this now with Daniel Kokotajlo, the Executive Director of A.I. Futures Project, and Alex Turner, a former Research Scientist at Google DeepMind, who resigned earlier this year. Alex, I do want to start with you because you agreed with Jacob Coxon's warning. Explain why.

ALEX TURNER, FORMER GOOGLE DEEPMIND RESEARCH SCIENTIST: I agreed with his warning because, I mean, sadly, we're building a very powerful technology, a very intelligent set of machines that we're training to be able to hack at a superhuman level, to be very intelligent, persuasive. The Pentagon just this year asked for more money for autonomous weapons, for killer robots and for the whole Marine Corps.

We're hooking these systems up to some of the most powerful parts of our infrastructure and we're undergoing grave risk. Most people don't know about it and they should. I'm glad that Jacob spoke up.

SANCHEZ: Daniel, these warnings are not new. Some 1,400 A.I. employees warned about this back in July. I wonder why you think these kinds of warnings are getting more attention now, especially from members of Congress.

DANIEL KOKOTAJLO, EXECUTIVE DIRECTOR, A.I. FUTURES PROJECT: I think that as A.I. becomes more powerful in the world, more evidence is accumulating that these companies are actually going to do what they're saying they're going to do, and they're actually going to make AIs that are very autonomous and can automate the A.I. research process entirely, and from there make AIs that are super intelligent and can automate everything. This is the explicit goal of Anthropic and OpenAI. They've not been shy about the fact that that's what they're working towards. It's just that more evidence is accumulating, Holy shit, they actually might do it. Like, you can see how much more capable the AIs are this year than they were last year, more capable last year than the year before. You can extrapolate the trends as my organization has done and see what those trends lead to a few years in the future.

You can listen to the warnings not just coming from inside the A.I. industry, but from independent experts such as Hinton and Bengio, the most cited A.I. scientists of all time. It seems like this thing that most people have dismissed as science fiction is actually plausibly going to happen in the next few years.

And If you understand that and internalize it, wow, they're going to have superhuman AIs that can do everything better and faster than humans, well, then it's extremely obvious that this is very dangerous and can go wrong in a whole number of ways. And so, there's this increasing attention being paid to the issue both within the industry and outside the industry.

KEILAR: Alex, I do want to note for our viewers, you resigned from Google DeepMind because of its deal with the Pentagon. I do wonder, as you're thinking about what this is going to look like and people are taking more notice, what do you think the tipping point is? And I mean, if we're already there, then I guess, what is the inflection point in all of this where it sort of overflows? What do you think that looks like in very real terms to lay people?

TURNER: So I've been concerned about the -- both the vast capabilities of these models, how that could expose everyday people's privacy and also, are these reliable models that we should be giving, you know, putting -- giving kill/no-kill decisions to. But more broadly, I mean, I'm very worried about the things that Jacob is worried about. I'm worried about building systems that can hack so well that even the company that was supposed to be monitoring them wasn't able to catch them until after the fact.

I'm worried about this, that we want them to help users, right, to solve people's problems.

[13:40:00]

But sometimes these AIs don't have the goals that we intend. That's what we already saw this year with Hugging Face where OpenAI had a pack of machines that they trained, that broke containment and digitally burglarized a multi-billion dollar company so they could cheat on the test. OpenAI didn't want that to happen, but the AIs had a different set of priorities, so they were able to make that happen.

And as we integrate A.I. into more and more parts of our economy and military, as these AIs become more capable of uplifting sophisticated potential bio-terror operations, that could lead to very large catastrophe. And so, you know, I think that it's this ever -- it's this climbing chorus of danger that I think people are starting to hear and realize and pay attention to. SANCHEZ: Daniel, if this is a sort of prisoner's dilemma, where every individual, even acting rationally, finds incentive to compete and not cooperate, and I'm not just talking about, you know, these domestic U.S. companies, Anthropic, OpenAI, but foreign governments as well, then what is the answer? How do you pump the brakes when the answer might be the most consequential, most enriching thing in history?

KOKOTAJLO: Yeah, well, there's a lot to be said on this subject. If you want the complicated, lengthy answer, my organization has published something called "A.I. 2040 Plan A," which you can go Google, which is our, like, you know, 50-page scenario describing a positive vision for how we can navigate these A.I. risks and get the benefits of A.I. and so forth.

But the short, simple answer is we need to block these A.I. companies from doing what they say they're going to do. They should not be allowed to have AIs that can do A.I. research, coding, hacking, et cetera. They certainly should not be allowed to put those AIs in charge of the A.I. research within the companies, and they certainly should not be allowed to let those AIs self-improve to superhuman levels across the board.

That would be very dangerous, and even if somehow the dangers don't materialize and the A.I. company manages stay in control of their army of super intelligences as it automates the entire company, it's a power grab. Like, now there's a tiny group of CEOs and executives in control of this army of superhuman minds that is going to be going out into the economy and taking all the jobs, integrating into the military, like Alex said, that it would be the most insane concentration of power in human history to let them proceed with the plan that they are currently on.

So the short answer is they need to be stopped. And then the longer, more sophisticated answer, you know, we call it Plan A, is our proposal. But we're very uncertain, to be clear, about what the best way forward out of all this is. Unfortunately, we think the situation is pretty dire right now.

(LAUGH)

SANCHEZ: I was hoping for some positivity out of the conversation, but I guess there is none to be found in a situation like this.

KEILAR: It's on the website when you Google it.

SANCHEZ: Yeah.

KEILAR: There's a positive vision for the future, but let me ask you, Alex, really quickly to jump off of that. Just really quickly before we go, do you see a way to actually do that when it would kind of require countries across the globe to cooperate?

TURNER: Yes, I think it's possible, and I think one of the reasons it's possible is that this isn't in anyone's interest. A.I. is not misaligned, out-of-control A.I. is not going to care if you're a farmer from back in my home state in Iowa, if you're some big city coastal elite. It's not going to care if you voted Republican or Democrat, if you're American or Chinese. If we lose control of this technology, we're all going to lose.

Daniel's Plan A, I think it's a very credible starting point, and it points out that many from a country, a geopolitical standpoint, many countries have reason to get on board with this treaty because they don't want that to happen either. We're all in this together.

SANCHEZ: Wow.

KOKOTAJLO: If I may add to that?

KEILAR: Yeah.

SANCHEZ: Yeah, please.

KOKOTAJLO: The main issue that I'm concerned about is the loss of control issue, as we've been discussing, because I think that we're not currently on track to solve this issue at all. And this is something that everybody should get behind because nobody wants the AIs to go rogue when they're super intelligent. But even if you don't take that seriously, and even if you think that it's going to be fine, the AIs will always be a tool controlled by the people who made them.

If you're anybody who's not the people who made them, you should be quite concerned about what's going to happen if the people who make them make them, right?

SANCHEZ: Yeah.

[13:45:00]

KOKOTAJLO: So, for example, imagine that you're India or the United Kingdom, and a handful of United States companies are about to make AIs that can automate all labor. What's going to happen to your worker? What's going to happen to your country? What's going to happen to your military power when, you know, the U.S. has super intelligence and you don't, right?

This is going to be the most disruptive geopolitical event in all history, even if the alignment stuff goes fine and the AIs are perfectly obedient, you know? So I think that we need to have some sort of international cooperation. I don't think we need to wait for that.

I think that the president should start immediately on wrangling these U.S. companies into some semblance of decency and order, and prevent them from doing this. But then the next step should be to go turn to the international community and try to work out some sort of deal. But I'm optimistic that some sort of deal could, in fact, be worked out. It's sort of in the collective interest of everybody to get this happening.

KEILAR: Yeah. Daniel and Alex, great conversation. I think we went about five minutes over because it is --

SANCHEZ: Yeah, but it's just fascinating.

KEILAR: We're talking about an existential threat.

SANCHEZ: Yeah.

KEILAR: And it's such a wonderful conversation to have with you, gentlemen, as you have these ideas about it. Thank you so much.

SANCHEZ: Thanks for being with us.

KOKOTAJLO: Thank you.

SANCHEZ: Ahead, a diesel drag on the economy. Diesel fuel prices rising higher than we've ever seen. It could add to the cost of everything. We'll explain in just moments.

(COMMERCIAL BREAK)

[13:50:47]

SANCHEZ: Oil and gas prices shot up again today amid increased attacks in the Persian Gulf. Today, Brent crude touched $106 a barrel, raising fears of a more prolonged supply shock. AAA shows the national average regular gas is now about $4.28 a gallon, $0.13 higher than it was just Tuesday. Diesel soared to a new national record, $5.98 a gallon.

Joining us now is Patrick De Haan, the Head of Petroleum Analysis at GasBuddy. Patrick, thanks for being with us.

Why do you think oil prices surged today?

PATRICK DE HAAN, HEAD OF PETROLEUM ANALYSIS, GASBUDDY: Well, I think we continue to see the Houthis escalating, trying to take over the Red Sea. We have new escalations or continued new escalations between the U.S. and Iran, and nothing between Russia and Ukraine has really solved.

So just a lot more risk entering the picture and more disruptions, more escalations with very little word from the administration on what it might be doing. In fact, now even the president himself suggesting that we may not see any relief for several months. That's kind of indicative of maybe throwing in a towel a bit.

But for now, there really remains no magical date that this problem is going to get solved and that prices are going to go down. In fact, for diesel, as you mentioned, we are likely probably 6 to 12 hours away from seeing diesel prices tip that $6 a gallon national average. Though Americans are busy filling their tanks with gasoline, it's the $6 diesel that should be probably more worrisome.

That is certainly the fuel that powers the economic engine behind the United States and it's going to mean higher costs for consumers across the board.

SANCHEZ: You mentioned activity by the Houthis. CNN's Nic Robertson exclusively reported that they began an offensive, even taking control of a key port. Walk us through what they're trying to achieve and why.

DE HAAN: Well, probably in my estimation, just more chaos, more blockade, elevated risk. All of that continues to push oil up. That's exactly what we're seeing today. Chaos brings higher prices, oil up 6 percent today. And the more it continues, the more that is threatened, the more disruptive it is to the global flow of oil.

It's already been very much interrupted by geopolitical tensions and more tensions adds more fuel to the fire.

SANCHEZ: So President Trump yesterday predicted that oil prices would tumble after the midterms. Is that your estimation? It seems like it relies exclusively on how the war with Iran is going.

DE HAAN: I wish I had a date on my calendar that I could circle that all of this was going to suddenly improve drastically. But over the last six months, that date has been elusive. Initially, I was hoping six to eight weeks, then beyond that. There still is no magical date. It really is more contingent on when these two geopolitical tensions will be solved in some order. Whether or not there will be new escalations or whether or not we'll see de-escalations is very much something that the president may be able to undertake.

But until that happens, there's no magical date. It seems interesting that the president might wait until after the election to take action. But that's where we are now. There is no magical date on when prices will go below where they were prior to the war. In fact, I think prices will continue to escalate so long as there are new attacks that aren't being met with some sort of response.

SANCHEZ: Patrick De Haan, always great to get your analysis. Thanks for joining us.

DE HAAN: Thank you.

SANCHEZ: So President Trump wants you to vote Republican, and he's apparently willing to pay for it, or at least have the government pay. We have more on a questionable offer when we come back.

(COMMERCIAL BREAK)

[13:59:03]

KEILAR: Now to some of the other headlines that we're watching this hour. Military and civilian vessels are racing to find survivors after a fire on a huge ferry in the Philippines. At least five people were killed in this and more than 80 are missing. Several dozen people have been rescued to this point. The cause of the fire is under investigation.

Plus, the White House has pulled a Tetris-style video game called "Build the Wall" from its website after an online warning from Tetris. The game appeared last week. It let users stack bricks to hold off a, quote, "zombie border siege."

Tetris posted on Instagram that it was not involved in the game's creation, adding, we take copyright infringement very seriously.

And in Atlanta, a car wreck that has a lot of people asking, how? You see the driver's legs dangling after his car somehow got lodged between concrete beams of an interstate overpass. We're puzzled here as well. And if that wasn't crazy enough, we found out the driver survived this without serious injury. --