Return to Transcripts main page

CNN News Central

Sources Say U.S. Military Had Close Call After Using A.I. For False Intelligence Report; Trump to Make Announcement on Prescription Drug Pricing; Holdout Juror in Clancy Murder Trial Speaks Out; Patrick Clancy's New Wife Defends Him Amid Conspiracy Theories. Aired 2-2:30p ET

Aired September 18, 2026 - 14:00   ET

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

(COMMERCIAL BREAK)

[14:01:52]

BORIS SANCHEZ, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": A CNN Exclusive, how the Pentagon nearly launched a military operation all because of an intel report created with the help of artificial intelligence. One official saying this almost started a war.

And priced out, as costs keep rising for Americans and Republicans worry about the midterms, President Trump makes an announcement this hour on health care.

BRIANNA KEILAR, CNN CO-ANCHOR OF "CNN NEWS CENTRAL": And no doubt, the holdout juror in the Clancy murder trial telling his side of the story and explaining why he refused to go along with the other 11 members of the jury.

We're following these major developing stories and many more, all coming in right here to "CNN News Central."

SANCHEZ: We start this hour with the CNN Exclusive Breaking News, sources revealing that the U.S. military almost launched an operation to intercept a Chinese vessel based on erroneous information provided by A.I.

KEILAR: One source says the intelligence report that circulated last spring was, quote, "entirely false" but almost started a war. CNN's Zach Cohen is here with more on this exclusive reporting. Zach?

ZACHARY COHEN, CNN SENIOR NATIONAL SECURITY REPORTER: Yeah, guys, sources telling me and our colleague, Katie Bo Lillis that an intelligence report was circulated within the U.S. military earlier this spring while the war with Iran was ongoing, and it immediately raised alarm within the Pentagon as it pointed to a Chinese vessel that was transiting through the Middle East and it indicated that that vessel was carrying components of a nuclear weapons program.

Now, obviously, the U.S. military reacting Very quickly to this intelligence report once they saw it, were preparing potentially to board this vessel. U.S. military aircraft were in the sky at the moment that officials dug a little deeper into the report itself and found that it was generated with the help of A.I. and that a chatbot that was used by a special operations analyst to put together the report had misidentified what this ship was actually carrying.

And this is an example, our sources say, of a very near miss, a potential close call for the U.S. military that could have led to a direct potential confrontation with China if the U.S. had followed through on this operation based on what we now know to be faulty intelligence generated by A.I. and contributed to by A.I.

And so, sources talked to us about how this is indicative of a broader problem. This is not the only incident we're told where A.I. has been used to generate intelligence reports, intelligence reports that some of which had been not vetted fully by the time that they were put into operational use and acted upon. One source saying that this is a growing and increasingly prevalent thing, the use of A.I., particularly when it comes to the use of A.I. in targeting operations.

Source said, quote, "A.I. in targeting is definitely something that is ramping up and there is no real guidance for how having a human in the loop will prevent civilian casualties or fratricide or in this case, a potentially direct confrontation with the people on that Chinese vessel."

[14:05:00]

So ultimately though, this operation was stopped once officials realized that the underlying intelligence was faulty. Again, though, once we're saying that the intelligence report itself was, quote, "entirely false" and that, quote, "it almost started a war," just speaking to the stakes potentially that were in hand here.

SANCHEZ: And to those stakes, Zach, what is the Pentagon's underlying strategy when it comes to artificial intelligence?

COHEN: Yeah, Boris, it's interesting because the Pentagon, under Pete Hegseth, the Defense Secretary, has been very A.I. forward since he took over that job last year. And he's even laid out in a memo a strategy for how he wants to see the U.S. military incorporate A.I. into pretty much every facet of what it does. And that includes targeting, that includes even more mundane tasks like budgeting.

And so, Hegseth laid out this strategy earlier this year, really saying that he wants the Pentagon to be experimental in how it uses A.I. Take a listen to what Hegseth said when he was introducing this strategy.

(BEGIN VIDEO CLIP)

PETE HEGSETH, UNITED STATES SECRETARY OF DEFENSE: This strategy will unleash experimentation, eliminate bureaucratic barriers, focus on investments and demonstrate the execution approach needed to ensure we lead in military A.I. and that it grows more dominant into the future.

In short, we will win this race by becoming an A.I.-first warfighting force across all domains.

(END VIDEO CLIP)

COHEN: So Hegseth and other proponents of A.I. say that it will allow the military to move faster, make decisions in the battlefield in real time. Obviously, we've heard from sources that say A.I. will help you move faster, but ones were saying, quote, "move faster towards a bad decision."

SANCHEZ: Zach Cohen, really a stunning quote and great reporting there. Thank you for being with us.

For more on all of this, let's talk to Alex Turner. He's a former Research Scientist at Google's DeepMind A.I. lab. He resigned over its partnership with the military.

So I wonder, Alex, when you hear about this reporting regarding the faulty A.I.-assisted intelligence report that nearly triggers a U.S. military operation, what is your response?

ALEX TURNER, FORMER GOOGLE DEEPMIND RESEARCH SCIENTIST: I fear that we are going A.I. forwards off a cliff right now because the speed with which we're integrating these systems, I don't think it's responsible. I think there's a lot of wins we can have here by pacing and moving forward appropriately.

When I was at Google DeepMind, I drafted dozens of pages of proposed contract language, ran them by leading experts in military, A.I. integration law, surveillance law that were meant to help Google check that there would be standards on this kind of use.

Executives ultimately didn't want to go for it. And I fear that a lot of people from Congress to people at these companies are kind of pretending that we're not having issues here. We need structure. It's not just enough to get a smart A.I., put it in the system and hope it works.

SANCHEZ: What were you proposing that these executives hesitated to adopt?

TURNER: I think basically, Google was facing, I think, a lot of pressure politically to go forward with this deal. I'm not opposed to A.I. in the military per se, but I proposed a set of red lines that would have allowed a lot of common sense proposals, common sense applications, leaving it to the military to dictate how they run it, but making sure that we would be engaging in applications, selling A.I. for applications that had processes.

SANCHEZ: Help us understand what that means. Did you get any sense that they were concerned that the Pentagon could potentially use this information in a way that would ethically compromise the work of Google?

TURNER: I don't think I can really elaborate on that topic.

SANCHEZ: Sure.

TURNER: I had a couple of communications with them. But ultimately, they decided not to move forward.

SANCHEZ: Understood. So across the military and the intelligence community, officials have pushed to weave A.I. in just about every facet of their work, from analyzing volumes of raw intelligence on how the U.S. collects and selects targets for strikes to just mundane bookkeeping and budgeting supply chains.

I wonder what you see as the broader risk of incorporating A.I. so broadly and further going into what the defense secretary described as an experimental use of the technology.

TURNER: It is a very new technology. It's a very powerful technology. We heard just now about this potential great power conflict that mistaken information almost caused.

[14:10:00]

And so I think -- I think we don't need to speculate too much about the potential harms from insufficient structure, which it sounds like happened here.

SANCHEZ: Sure. So we're also learning that a group of independent hackers used Anthropic's Claude to find a way in to an OpenAI employee's ChatGPT account. This was first reported by the Wall Street Journal. What questions does that raise for you?

TURNER: I think it raises who else has been in OpenAI's key code base, accessing their many insights on how to train these systems, stealing IP, corporate espionage. It really makes me wonder. You know, this is not a state-sponsored elite hacking group. It's a couple of guys who probed their software for a little while.

So, I'm very concerned and I'm wondering, you know, does OpenAI have any idea who is in their code base?

SANCHEZ: Wow. The question of how companies train their A.I. models is a separate and important one. I spoke with the godfather of A.I., the so-called godfather of A.I. last night. Here's what he shared with me.

(BEGIN VIDEO CLIP)

GEOFFREY HINTON, COMPUTER SCIENTIST & PROFESSOR EMERITUS, UNIVERSITY OF TORONTO: The big companies are trying to make them more and more intelligent. We should also be trying to make them more and more well- behaved. And the current way you make them well-behaved is you train them up on lots and lots of data, and they have lots of terrible tendencies, and then you try and get them to repress those terrible tendencies. That isn't a good way to build a reliable system.

(END VIDEO CLIP)

SANCHEZ: That, of course, Geoffrey Hinton. What do you think of how he's describing the way that AIs are trained and the way that we try to get some form of compliance from them? TURNER: I agree with his broad concerns. We are training these AIs. We're kind of growing them, more like. We use this ingredient called compute, computing horsepower, to grow AIs. And the more compute you use, the smarter they are. No one really knows how they work.

And in particular, we've already observed real mistakes and misalignment with what the AIs actually do versus what the people who built them, their developers, what they intended. And so this is a misalignment risk. We saw it with the Hugging Face hack of a multi- billion dollar company by OpenAI's agents. And this is the dumbest A.I. is ever going to be. These systems are going to get smarter, and the consequences for failure will get more severe.

SANCHEZ: That is such an important point. This is the dumbest that A.I. is going to be. Alex Turner, thanks so much for sharing your expertise.

TURNER: Thank you.

SANCHEZ: Brianna?

KEILAR: Minutes from now, the White House says President Trump will be making a health care announcement, one that aims to lower the cost of prescription drugs for millions of Americans. Let's go to CNN's Kristen Holmes for more on what the president will say. Kristen?

KRISTEN HOLMES, CNN SENIOR WHITE HOUSE CORRESPONDENT: Well, Brianna, we're told by the White House that President Trump is going to make an announcement that all 50 states, as well as Puerto Rico and Washington, D.C., are going to participate in a model Medicaid program in which they will get drugs at a discounted rate.

Now, this is all part of the Most Favored Nation program. This has been a big push for President Trump and something that he's been campaigning on during the midterms. And essentially what that is, is we've seen these announcements kind of piecemeal. But overall, more than two dozen of these Medicare -- medical and drug companies have made agreements to essentially provide medications to Medicaid as part of a program on prices that are based at the lowest cost available to any developing nation.

Again, this is going to be a big part of President Trump and the administration's push as they try to get out there with midterm messaging. One thing to note, this White House official did also tell us that there are going to be governors from various states in the room with him.

Another thing to point out, we haven't really seen President Trump available for questions in a few days. So this will be an opportunity to also ask him about recent events.

KEILAR: All right, Kristen Holmes, live for us at the White House. Thank you.

Still to come, J.D. Vance campaigning in Iowa with 46 days till the midterms. (COMMERCIAL BREAK)

[14:18:46]

SANCHEZ: Republicans have a new pitch to voters with the midterms just 46 days away. Here's Vice President, J.D. Vance campaigning in Iowa just minutes ago.

(BEGIN VIDEO CLIP)

J.D. VANCE, (R) VICE PRESIDENT OF THE UNITED STATES OF AMERICA: Always tell you, ladies and gentlemen, whenever you hear me talk during this midterm campaign, I'll say Rome was not built in a day because we know there's a lot of work to do.

(END VIDEO CLIP)

SANCHEZ: CNN's Jeff Zeleny joins us now. Jeff, similar to what Vance said on a podcast this week, give us another chance. Is that what voters are going to think about when they head to the polls?

JEFF ZELENY, CNN CHIEF NATIONAL AFFAIRS CORRESPONDENT: Well, we'll see, but I mean it certainly is an acknowledgement here that, look, we understand that things are not going as well as we would maybe like, but what's so interesting about it is it's totally discordant from what President Trump is saying. He repeatedly says America has the hottest economy in the world. He repeatedly says everything is fine.

But the vice president, this is now the second or third time in a row in his recent campaign events, as he's really been stepping up his travel, he's acknowledging that the economy is bad. I did not hear him mentioned specifically about the high gas and diesel prices, but of course, everyone in the audience who drove to this event is well aware of that.

But I think making comments like that Rome wasn't built in a day, we're working on it. But the give us a chance, he did not repeat that and I think that was probably by design because that is almost certainly to be used by Democrats against them.

[14:20:00]

Some would say they've had a chance, but I think him acknowledging that things are not as good as they can be certainly is notable, but again, different from the president.

SANCHEZ: The list also long of the issues that President Trump vowed during 2024 would be resolved on day one with Republicans walking through the door. Notably, there was an absence here at that rally. The Iowa Republican Senate candidate, Ashley Hinson, decided not to go. She had a previous obligation, I think was the official term, but just given where things are in the state of politics right now, you would think that if she found it enormously helpful to be alongside J.D. Vance for her campaign, she probably would be, right?

ZELENY: You might think so. He is the Vice President of these (ph) United States, but I think back to so many midterm elections and election days when scheduling conflicts arise for Democratic incumbents when the president -- or Republican incumbents, when the president is not doing so well. But take a listen to J.D. Vance. He actually praised her, but also called her out a bit.

(BEGIN VIDEO CLIP)

VANCE: First, I want to make it a shout out to an amazing public servant. I know she couldn't be here today, but are we going to elect Ashley Hinson, the next United States Senator for the great state of Iowa?

(END VIDEO CLIP)

ZELENY: So certainly a shout out there, but also drew attention to the fact that she was not there. I mean, the vice president was in Iowa talking about the governor's race, but what's so interesting about Iowa, this is a state that he and President Trump won by 13 points just two years ago. Now, it is a wash in competitive races, the governor's race, the Senate race, three competitive House races.

And by my count, he also was missing some other members of Congress from Iowa, two of which switched their votes this week on the Iran War Resolution in the House. So at this point of the campaign, with six weeks left, Republicans across the board are doing what they can to show some independence.

The question is, will voters sort of see that's enough or not? But very notable, who's campaigning where and who's not showing up?

SANCHEZ: Jeff Zeleny, thank you so much for that report.

So it's a case that got national attention. Patrick Clancy and his new wife are now giving their first interview since the murder trial for his ex-wife, Lindsay Clancy, ended with a hung jury.

Plus, we are hearing from that lone holdout juror. Those details after a quick break.

(COMMERCIAL BREAK)

[14:27:02]

KEILAR: Today marks two weeks since the Lindsay Clancy murder trial ended with a hung jury. The panel couldn't reach a consensus on whether Clancy was criminally responsible for the deaths of her three children. Now her ex-husband, Patrick, is speaking out for the first time. Here's some of what he told CBS' "60 Minutes" about how he's coping with the deaths of their kids.

(BEGIN VIDEO CLIP)

PATRICK CLANCY, LINDSAY CLANCY'S FORMER HUSBAND: I talk to them.

UNIDENTIFIED MALE: What do you say?

CLANCY: Help me. All the time.

UNIDENTIFIED MALE: Do they help you?

CLANCY: I like to believe they do. It's a way I stay connected to them.

(END VIDEO CLIP)

KEILAR: CNN's Brynn Gingras is with us now. Brynn, what more can you tell us about this interview?

BRYNN GINGRAS, CNN U.S. NATIONAL CORRESPONDENT: Yeah, hey, Brianna. So Patrick Clancy actually gave an interview prior to the trial actually happening, but that was an in-person on paper. This is the first one that's going to be happening on T.V., and the first one after that mistrial was declared, like you said, just two weeks ago.

So a lot of eyeballs will be on this. Of course, he and his now current wife were really the subject of unfounded conspiracies that circulated everywhere, on social media, online, during this trial. And so you can imagine, he's probably going to be speaking to that, but also his wife, his current wife, is also going to be part of this interview with "60 Minutes."

Take a listen to some of what she has to say.

(BEGIN VIDEO CLIP)

DR. RACHEL DANIS, PATRICK CLANCY'S WIFE: I don't understand how the story has been so twisted, but I would hope people know that Pat is, as I said before, the most selfless father and supportive partner. And I was just a girl that met a guy running in Central Park on a really cold Saturday. And I don't think there's anything wrong with that.

(END VIDEO CLIP)

GINGRAS: Very clear. They're trying to set the record straight, as I said, because they were the subject of so many unfounded conspiracies. Now, they are speaking out. That's airing this weekend. That is ahead of Lindsay Clancy's next trial date. That is for a hearing on September 29th. That's where we might hear if the prosecution is going to decide if they'll retry this case, Brianna.

KEILAR: And Brynn, we're also hearing from the lone holdout juror whose vote forced the trial to end in a mistrial. What's he saying?

GINGRAS: Yeah. Remember, we've heard from some other jurors, including the four women in this case, who all sort of pointed the finger at this single juror, who said really he's the reason that we went to a mistrial, that he wasn't following the rule of law, that he actually expressed doubt in that juror deliberation room.

But he is coming forward through his attorney and setting the record straight as well. He posted on X, or at least his attorney for him posted on X this, "I didn't have any doubts. As I tried to explain different possible theories during deliberation, I kept getting cut off as if I had doubts based on the evidence presented."