Return to Transcripts main page

CNN News Central

Prediction Market Trading Tops $750 Million on Midterm Outcomes; Elizabeth Holmes Approved for Halfway House Next Year; Trump Hosts Chinese Leader for Summit in Washington; White House Tells Judge Media Access Has Been Fully Restored; Trump Rejects Calls for A.I. Guardrails; OpenAI Agent Hacks Into Australia's National Healthcare System; Nor'easter Gains Strength as It Takes Aim at East Coast. Aired 1:30-2p ET

Aired September 24, 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]

MARSHALL COHEN, CNN SENIOR REPORTER: -- election rigging and ballot stuffing. I spoke about this with the top election official in L.A., Dean Logan. Here is what he said was his message to the prediction companies.

(BEGIN VIDEO CLIP)

DEAN LOGAN, LOS ANGELES COUNTY CLERK: Don't do this. Stay out of this particular field. There are plenty of other areas for you to do what you do, but you don't really belong here. It's not appropriate because it does strike at the fundamental purpose of an election, and anything that is creating a lack of public confidence in that is destructive.

(END VIDEO CLIP)

COHEN: So they know this stuff is legal, but they kind of wish that it wasn't. And of course, for full disclosure, CNN has a partnership with Kalshi. We use its data to cover major events, but they don't influence our coverage, and we don't trade on prediction markets.

BRIANNA KEILAR, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": A candidate checking their odds every hour?

COHEN: Yeah.

KEILAR: They must read their comments too, right? Stay out of the comments. Stay out of the odds.

(CROSSTALK)

COHEN: (Inaudible) that's grass (ph).

KEILAR: That's right. Ground yourself. There are other ways to spend your time. OK, how are prediction companies responding?

COHEN: They are pushing back against a lot of these critiques, and in fact, they've made the case that they're going to increase turnout. They say they're bringing people into the political process, especially young Gen Z traders that love these markets and maybe wouldn't have been engaging in politics before.

Let me read you some reaction from Kalshi. Their Head of Politics Growth said this, quote, "Polls are backwards looking. They're expensive to run. They can be biased, but markets are always live. They're 24/7."

So they're making the case these are better than the polls. This is a public service. And then Polymarket said we provide transparent, accurate, real-time information. They said that their markets are an antidote to the disinformation poisoning public dialogue.

KEILAR: Maybe that candidate agrees with them, right?

BORIS SANCHEZ, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": Probably. Yeah.

KEILAR: All right. Marshall, fascinating. We'll see. Time is going to tell us whether this is -- whether they're right or not.

COHEN: Yep.

KEILAR: Marshall Cohen, thank you.

COHEN: Thank you.

KEILAR: Next, convicted fraudster Elizabeth Holmes could end up serving less than half of her 11-year prison sentence. Find out why right after this.

(COMMERCIAL BREAK)

[13:35:51]

KEILAR: President Trump rolling out the red carpet for Chinese leader, Xi Jinping, greeting him in a rare plane-side welcome ceremony. Later saying Xi looked, in Trump's words, strong, vibrant, and fit. The three-day visit full of pomp, but it's unclear how much substance will actually come from this. There are major questions unanswered about the A.I. race, about China's support of Iran, and the fate of Taiwan.

We're joined now by five-time U.S. Ambassador, Christopher Hill. He also served as former Assistant Secretary of State for East Asian and Pacific Affairs. Ambassador, thanks for being with us.

As I noted, the expectations for this visit are pretty low. But should they be? Is this a missed opportunity at this critical time in U.S.- China relations?

CHRISTOPHER HILL, FORMER ASSISTANT SECRETARY OF STATE FOR EAST ASIAN AND PACIFIC AFFAIRS: I think the problem is the summit seems to be the event in and of itself. That is, they're not doing the hard work that goes on between summits. You know, summits are a time where you pull a rabbit out of a hat, but someone has to spend some time stuffing that rabbit down the hat. And we don't see a lot of that.

I think, secondly, there's always been this issue where the U.S. thinks China can do more than it's doing, and the Chinese always say they can do less than they can do. So it's always a, you know, difficulty of finding where the Chinese can be helpful. And these problems are really mounting. These international problems are mounting for this administration, only five weeks away from our elections.

And so there's always this tendency to go or to think there's some kind of one-stop shopping with the Chinese, and there isn't. So I would set the expectations, they should be fairly low for this summit.

KEILAR: Yeah. And that's what we keep hearing. How does the president need to handle the issue of Taiwan here? Even as you say, expectations are low, if he doesn't want to rock the boat with China, but he needs to send a message that the U.S. is still very committed to Taiwan.

HILL: You bet, you bet. And it's a very delicate process, and I'm not sure our president is really set up to do delicate processes. But, I mean, he needs to show that the United States has an abiding commitment to making sure that Taiwan is able to mark its own future. And you know, it's going to be difficult with a China who feels that they have the leverage here.

So I suspect the end game will be somehow that the president will not cancel and not announce this arms package. I think it'll just be kicked down the road a little longer. You know, the arms package would take, you know, months, maybe years to be actualized.

So I think the president will probably try to satisfy the Chinese by saying, look, I'll hold off on the arms package. So these are tough issues for the president and he's looking for some real success here, and it's going to be tough going.

KEILAR: So Trump portrayed A.I. as a way to promote peace for decades to come while he was standing with Xi. This morning he posted, superintelligence will be a big topic of discussion, but I want to leave it exactly where it is. That is China's position also. Our guardrail is the DOJ.

He's saying this as his cabinet is discussing an A.I. hotline with China. What do you think about where this stands?

HILL: I think the A.I. hotline is a good idea, not that it solves much, but it does show some modicum of trying to work with the Chinese on this very difficult problem. I think the problem is that even within the U.S., you've got a lot of sharp disagreement of people. The president knows the importance of A.I. to our stock market.

At the same time, I think even the tech CEOs are expressing some concern about the degree to which it's moving along. As for renaming A.I. to S.I. or whatever, nice little marketing thing, but hardly the point right now.

KEILAR: Ambassador, I do wonder how you're looking at this White House court battle with the press, which includes CNN, of course, unfolding as Xi is in Washington and Chinese state media is at the White House. [13:40:00]

HILL: Yeah, it's kind of, frankly, crazy to say that the Chinese press is able to report on this and ours is not. And you know, I don't think I have to tell any American citizen the importance of the Fourth Estate to our country.

And my goodness, I hope we get this resolved very quickly and firmly, because, you know, we're an open country. And we need our press able to report on issues no matter what they 00 whether you like what they say. This is just -- this has to be addressed. And I hope every American is as frankly outraged as I am on this.

KEILAR: I also want to ask you about CNN exclusively reporting last week that the U.S. military was on the verge of boarding a Chinese ship. I mean, this got very close to spiraling out of hand after A.I. was used, foreign intelligence report that turned out to just be false, very false.

HILL: Yeah, yeah.

KEILAR: I wonder, I mean, whether or not that comes up or not. I wonder the timing of that. Does that make the U.S. look stupid going into this meeting?

HILL: You know, I think we need to do a better job of preparing ourselves for these issues of communicating with people and trying to reduce this mountain of trust -- of mistrust that has grown up between the U.S. and China, and the U.S. and the rest of the world. I think we have a lot of work to do. And these kinds of issues, when they come up, usually you have means to address them before they become an international crisis.

I mean, the idea of a U.S. boarding a Chinese vessel, I think, would really reverberate around the world. It would have really severe repercussions for what we really need to do to work with China. We need some understanding going forward in the Middle East. We need some understanding of how to manage the issue of Iran. We need some understanding, especially when we talked about North Korea.

But I mean, there are a lot of concerns there. And yet, here we are kind of lurching from one crisis to another because we don't have adequate planning mechanisms and we certainly don't have a platform of communication to deal with these things.

KEILAR: Ambassador Christopher Hill, thank you so much for joining us.

HILL: Thank you very much.

KEILAR: Just a day after top A.I. CEOs told world leaders at the U.N. that more guardrails are needed, President Trump says, nope, what's in place is enough. Who agrees with him? We'll talk about it next.

(COMMERCIAL BREAK)

[13:47:03]

SANCHEZ: We're learning about a new alarming incident involving A.I. We have our first known case of an artificial intelligence agent managing to hack into a government network. Australia's prime minister revealed that an apparent rogue OpenAI agent hacked into the country's health care database when it was looking at a research about health care spending. The agent then decided to find answers in areas it had no authority to go into. The breach happened back in June. Australia though says it wasn't notified about it until earlier this month, a timeline it calls unacceptable.

OpenAI's Sam Altman and Anthropic's Amodei spoke about the risks of A.I. at the United Nations Security Council yesterday.

(BEGIN VIDEO CLIP)

SAM ALTMAN, CHIEF EXECUTIVE OFFICER, OPENAI: We need common standards so countries can compare evidence, verify compliance and have a shared language and understanding about what is happening. We need accurate and speedy incident reporting, classification and reporting protocols, so the world can learn from failures before they become catastrophes. And we need secure channels among governments, critical infrastructure operators and technical experts to share emerging vulnerabilities and new threats.

DARIO AMODEI, CHIEF EXECUTIVE OFFICER, ANTHROPIC: I believe that this is the most important global security issue facing the world today. For all the differences of the people in this room, we must put aside those differences in order to confront this global opportunity and global threat that is being presented to us at the same time.

(END VIDEO CLIP)

SANCHEZ: While A.I. leaders call for an international framework, President Trump posted the following on social media this morning ahead of his meeting with Chinese leader, Xi Jinping, quote, "Our guardrail is the DOJ."

Joining us now is Daniel Kokotajlo. He's the Executive Director of the A.I. Futures project. Daniel, thank you so much for being with us.

So OpenAI so OpenAI says that this breach of the Australian health care system occurred in June. They only learned about it in August though. What does that tell you about the kind of oversight that they have of these agents?

DANIEL KOKOTAJLO, EXECUTIVE DIRECTOR, AI FUTURES PROJECT: That tells you that it's not very strong. I think they are working to improve, but OpenAI has this strategy called iterative deployment which is this idea that we're going to learn how to deal with the A.I. safety problems by running into them as fast as we can and then fix them after the fact. And you know, we'll learn fastest by doing this sort of strategy.

And I've been saying for years, including while I was at OpenAI, that this strategy is going to be disastrous once the AIs are super intelligent.

SANCHEZ: You worked for OpenAI for two years on alignment and forecasting. I wonder if given your experience, you think the public can trust what Sam Altman and the company is saying.

KOKOTAJLO: Definitely not. I think that these companies are downplaying and low-balling the extent of the situation and the seriousness of the risks.

[13:50:00]

You know, the CEOs are now changing their tune and they're saying, yes, there are some risks and we need to have international coordination and regulation. But what they're not saying is that, basically, these companies need to be shut down until we can find a better way to build A.I.

You know, they do not have anything close to a solid way of training their AIs to have the desired traits and values, such as obedience and honesty. And yet, they are making these AIs smarter, faster. And they're planning to put the AIs in charge of research across their companies.

They are not at all ready to do this. This is exactly as dangerous as it sounds. And that's what they're not telling you. They're trying to make it seem like we can do sort of moderate reforms, like having some third-party evaluators come in or have some international phone calls between the presidents and so forth. We need something much less moderate than that.

SANCHEZ: Yeah. So, you mentioned the sort of red phone idea, an echo of the Cold War, when there was this mechanism between the United States and the old Soviet Union, in case there was some kind of mishap or potential miscommunication regarding nuclear weapons, they could contact each other.

Do you think that that's generally a good idea, as A.I. is undoubtedly on the table for discussion between President Trump and Xi Jinping during their meeting today?

KOKOTAJLO: Of course. I think all of that stuff is a step forward. It's just not going nearly far enough.

SANCHEZ: And when you hear President Trump say that our guardrail on A.I. is DOJ, and also the former FTC Commissioner, Lina Khan, sharing the belief that laws that are on the books are already in place to charge companies and their CEOs for any dangerous or unvetted or defective products. Clearly, it seems like you don't think that's going to be enough. I wonder if you think these companies understand the liability that may be associated with their products going into the future.

KOKOTAJLO: So the problem is that if they do recursive self- improvement and build superintelligence, then the catastrophic harms that result could be too big for the liability system to even exist afterwards to punish them for it. SANCHEZ: Wow.

KOKOTAJLO: Right? So if they are going to do this, they have to get it right the first try, rather than doing it, seeing how it goes wrong, and then being punished for it, and then adjusting their policies afterwards. So I think the thing that I would say to President Trump is, basically, you're being lied to by the tech industry plants that have been talking to you for so much.

They are downplaying the seriousness of the situation, and they're exaggerating the extent to which they understand what they're doing, and the extent to which they know how to make this safe.

SANCHEZ: I'm curious about what you would say to skeptics, because we've had these conversations about the dangers of A.I. and superintelligence, specifically spiraling out of control. And some of the feedback I've gotten from viewers and folks consuming these conversations is that it doesn't take into account the way that these are just algorithms and sort of bots that try to predict, through calculations, what to say in response to questions that you ask. It seems that they have a different view of the capacity of A.I. So I wonder what you would tell those skeptics.

KOKOTAJLO: A couple things. So both on the object level, I think that this idea of AIs as pattern recognizers or text predictors is kind of outdated. They're much more than that. That was sort of how the original Large Language Models were back in around 2020. But these days, they're trained to do so much more than that.

They're trained to operate autonomously for long periods in pursuit of goals, accomplishing diverse, challenging tasks in math and coding and hacking and so forth. So they're not just predicting text anymore. Also more philosophically, I mean, imagine there's this joke of someone looking at a tiger and saying, I shouldn't be afraid. It's just a bunch of biochemical reactions and atoms.

(LAUGH)

KOKOTAJLO: You know?

SANCHEZ: Yeah.

KOKOTAJLO: It's like, sure, yeah, in some sense, these things are just computers. But I mean, just look at what they're already able to accomplish. Look at the Hugging Face swarm -- the OpenAI swarm that broke out and attacked Hugging Face. Now, imagine something like that happening, but with much more powerful AIs that are much more numerous and that have been entrusted with running the company.

SANCHEZ: Yeah, it's just a big cat, a tiger after all.

(LAUGH)

KOKOTAJLO: Yeah.

SANCHEZ: Daniel Kokotajlo, thank you so much for joining us. Appreciate your time.

KOKOTAJLO: Thank you.

SANCHEZ: So there's a rare September nor'easter, and it is set to slam into several states, bringing floods, powerful winds and heavy rains. We've got the details on your forecast when we come back.

(COMMERCIAL BREAK)

[13:59:34]

SANCHEZ: Now to some of the other headlines we're watching this hour. Millions of people from North Carolina to New England are bracing for a rare September nor'easter. The storm is gaining strength and is expected to hit states with coastal flooding, heavy rain, and powerful winds starting as early as tomorrow night. Some areas have already seen strong gusts and flooding, so a lot to stay on top of there.

The USDA says that more than 167,000 pounds of meat are being recalled across the U.S. after being produced without proper federal inspection. Star Meat Delivery Inc. based in North Carolina --