Return to Transcripts main page

The Situation Room

Lawsuit: Top A.I. Companies Made Illegal Agreement On Slowdown; Rep. Jay Obernolte (R-CA), Is Interviewed About Trump To Appoint A.I. Czar But Rejects Calls For Regulation; Today: Mayor Mamdani To Meet With Pres. Trump In NYC. Aired 11-11:30a ET

Aired September 21, 2026 - 11:00   ET

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

(COMMERCIAL BREAK)

[11:00:47]

PAMELA BROWN, CNN ANCHOR: Happening now, decoding A.I., a new warning after Google's Gemini goes rogue. And President Trump is ready to create an A.I. force and name a czar for the technology. And then later, press power, CNN, and other major networks are now taking the White House to court after President Trump banned them.

Welcome to our viewers in the United States and around the world. I'm Pamela Brown. Wolf Blitzer is off, and you're in The Situation Room.

And we begin this hour with another fresh slew of headlines on artificial intelligence. This morning, we've learned more about a new lawsuit against several prominent A.I. companies. The plaintiffs claim that Anthropic, OpenAI, SpaceXAI, and Google all violated antitrust laws when they made a deal to slow the pace of A.I. development. We're going to have more on that in just a moment.

Also this morning, President Trump has announced plans to create a, "A.I. force and czar," even though he recently called fears of an A.I. apocalypse and the concerns around A.I. a hoax. All of this as he is set to meet with Chinese President Xi Jinping at the U.N. General Assembly, where A.I. will be a major topic of conversation. Treasury Secretary Scott Bessent says Washington and Beijing are already discussing potential safeguards for the technology, like alerting one another about potentially dangerous A.I. incidents.

(BEGIN VIDEO CLIP)

SCOTT BESSENT, TREASURY SECRETARY: We think that just like with any cross-border activity, that moving from opaque to more transparency between the number one and the number two A.I. powers in the world is very important.

(END VIDEO CLIP)

BROWN: All right, let's go live now to CNN Business technology editor, Lisa Eadicicco. Lisa, let's start with this new lawsuit against several prominent A.I. companies. What more are you learning? LISA EADICICCO, CNN BUSINESS TECH EDITOR: Yes, Pamela. So the interesting thing about this lawsuit here, really what's at the heart of the concerns here is that this call for a slowdown in A.I. development. There's accusations here that that could potentially cause harm for consumers who are paying subscriptions for services provided by those companies.

The thinking being here that the plaintiffs allege in this lawsuit that consumers, when they're paying for these subscriptions, they're doing so under the impression that the individual companies, whether that's SpaceX, Google, Anthropic or OpenAI, are making independent decisions about the future of their products and the pace at which to continue developing them.

And so the argument here is that this coordinated effort could kind of threaten the pace of innovation. And these companies, if there isn't as much of a competitive factor here because they're coordinating on a slowdown, they might not be as motivated to improve their products as quickly.

And I think, you know, this is an interesting wrinkle to the ongoing concerns we've been hearing around the future of A.I. and safety. And I think it kind of underscores the challenges that these companies face in really just continuing to maintain and push forward the products that some people are coming to rely on for work and other everyday tasks, while also doing so in a way that is responsible and safe and takes into account these new unexpected instances of A.I. models acting, you know, going rogue and acting in ways that they aren't expected to during training.

So I do think this is one to keep an eye on for sure. And it kind of underscores that broader issue.

BROWN: I know that there's obviously been these public calls for a slowdown from the A.I. leaders, but is there any indication they're speaking privately to reach some sort of deal as this lawsuit sort of indicates could be happening?

EADICICCO: So there have been some discussions around the formation of an industry standard. Right now, there is no broad standard. There is no formal federal regulation. So there have been a lot of concerns here about what needs to be done to kind of make sure A.I. advances safely. And, you know, a lot of the really important factors here, of course, the models themselves and how they're tested are really critical.

[11:04:56]

But I also think a key thing to think about here when we're thinking about the advancement of A.I. is not just the technology itself and what it's capable of, but the decisions made by these tech companies and these tech leaders when it comes to how do they secure their models? How do they test them? How do they train them? What kind of data are they using to train them?

All of these decisions that these companies make really do kind of play into this the trajectory of A.I. and how these broader concerns are handled. And I think that's what we're seeing now is that these companies are in a phase where they're trying to figure out what that looks like moving forward.

BROWN: All right, Lisa, thank you so much.

And with A.I. dominating the headlines recently, we want to take a step back and revisit just the basics of what we're talking about here. So we can hear the conversations happening now with more clarity and context. This is all part of the Situation Room's Decoding A.I. series. And we are joined by CNN correspondent Tom Foreman to break it all down. Let's start with this question that may seem a bit silly, Tom, but it's key to the whole debate. What is A.I.?

TOM FOREMAN, CNN CORRESPONDENT: What is A.I.? It's a really key question because many people don't really know. A.I. performs tasks that typically require human intelligence. So, for example, you can hit send on an email and reply to somebody. That's not A.I. But if you were to say to your A.I. agent, this software in your computer, look, I want you to read all my e-mail and prepare a summary for the day of all the people who sent me messages and what they were about, then you're getting closer to A.I. because it can analyze data. It can recognize patterns and create content.

So for example, it could say I get three or four e-mails every day from some guy, Steve, but he's never interested in those. So I'll push those down. Those aren't important. But I only get one from Pam every couple of weeks and it's really important. I can move that up and it has words in it that matter. So I can move that up. But this is really important.

Even as it's doing it, that simple task, it can make mistakes or produce false results because it does not know what it is doing. Seems like a human, but is really not. So in effect, if I put it in human terms, if you had a statistician come up with your medical care based on the statistics of medicine, he'd probably be right a lot of the time. He's not a doctor. He doesn't know what he's doing. He's just reading math. That's what this is doing.

So what this does is it can create text through these methods. It can create images. You've seen plenty of those. You've probably read plenty of that. Even if you didn't know you were, it can seem very human like. It can create video, which is very compelling. You may remember that video of Brad Pitt and Tom Cruise in a fight scene for 15 seconds. All made up. Didn't really happen. It can create audio. You may have heard it in some music.

Music is being made this way easily. And it can write software code. So it can create new computer functions out there for you. And who's building all of this? Look, the big names, some of them you mentioned a minute ago, Anthropic, OpenAI, SpaceXAI, the others here. These are just the really big ones. There are dozens and dozens and dozens of country -- companies doing that in this country alone. And abroad there are even more.

So basically a lot of technology companies are at least dabbling in this now, if not embracing it wholesale, Pam.

BROWN: And Tom, apart from A.I., people on both sides of this debate sometimes sound like they're speaking a different language. We know this technology is rapidly evolving. What are the key phrases or terms we need to understand?

FOREMAN: There are too many of them. But just listen to these, kind of get them in your head. It may help you a little bit. Generative A.I., things that create content out there. Agentic A.I., multi-step processes, problem solving. This is generative A.I .is I need a picture of a cat wearing a hat with a message next to it for someone's birthday. Make that for me. But agentic A.I. is I run a factory. I'm going to give you all the information about my supplies, my assembly line, my shipments. Find a way to make it more efficient and save me money. That's what it can possibly do there.

And large language model uses existing text to understand, generate language. So it reads, reads, reads, reads, reads, reads huge numbers of things. And then because it can recognize patterns, you say, give me a new novel by Dostoevsky. And it will come up with something that kind of sounds like that. Probably not as good as Dostoevsky. But it can do that. So they can generate that with large language models.

Beyond that, alignment, these are the big questions, Pam, in your early stories that matter here. Alignment and guardrails, alignment, making sure A.I. does what people intend and that it follows human values. It's like Asimov's rules for robots here. Is it doing what you want? For example, if A.I. were told to keep track of your health and help you with your health, as I mentioned earlier before, what if it determines that in some bigger picture of things that maybe you just shouldn't be eating because eating is making you not feel well. So we'll just stop you eating. That's a really simple version, but that's not what you intended. You said, help me find better foods. Don't just say don't eat.

[11:10:02]

And guardrails, how do you keep A.I. from operating completely on its own and doing something you don't intend? So what we're talking about here is different levels, artificial narrow intelligence, specific tasks. Well, summarize my email, help me do that. That's one artificial general intelligence, broader level, human level and super intelligence. This is one of the big ones.

Is it really capable of taking off on its own, attacking other companies, changing electrical grids, doing all sorts of things? That's what has people very worried, Pam.

BROWN: Yes. All right. Tom Foreman, thank you for helping us better understand all of this. And you'll be back tomorrow. And we're going to talk about the potential strengths and dangers of A.I. and how we can make it safer.

So to continue this discussion, we are joined by Republican Congressman Jay Obernolte of California. He is vice chair of the Congressional A.I. Caucus and has a master's degree in the technology. He's also the co-sponsor of a new bipartisan act to strengthen the oversight of advanced A.I. known as the FRONTIER Act. Congressman, thanks for coming on.

You really are probably the most authoritative voice as a member of Congress on this issue, given your background. I first want to start with just your reaction to President Trump saying that he wants to appoint an A.I. czar and create an A.I. task force, despite continuing to call the concerns over A.I., "a hoax."

REP. JAY OBERNOLTE (R-CA), VICE CHAIR, ARTIFICIAL INTELLIGENCE CAUCUS: Right. Well, we welcome the executive branch's engagement on this issue. I share the President's feelings that overregulation is something that we should avoid because it's going to affect our geopolitical competitiveness. And I think there's a balance that can be struck.

And we're trying to do our part in Congress by passing common sense legislation that imposes a light touch regulatory framework that balances those two needs, the need to protect Americans against the very real risks of the harmful use of A.I. and by malicious human actors against the risks of overregulation and the need to create a technology that allows our economy to flourish and all the good things that A.I. is going to do for our society to occur.

So we think that the FRONTIER Act achieves that balance. It's broadly bipartisan, and we're going to be pushing to pass it through the Congress this year.

BROWN: Tell us what the light touch means in your view, what this act does. And when you say human actors using it in the wrong way, I mean, you also heard concerns about A.I. not with the use of human actors going rogue and causing great harm to society that's not in alignment with human values. How would this tackle that?

OBERNOLTE: Right. Well, I think it's important to keep in mind that these recent incidents have been exposed through testing. And the models that we're testing in all of these cases were models that had their safety controls intentionally removed. So the companies were asking, as they should, what's the worst thing that can happen? And, you know, we're finding out what the worst thing that can happen.

So we're finding out a lot more about alignment, where we need to align the values of A.I. to human values. And we're finding out a lot more about the ways that we impose guardrails on A.I. For example, it's becoming increasingly clear that what A.I. developers call chain of thought monitoring is going to be required to ride herd on these advanced models.

But to answer your question, what we think the correct regulatory approach is, is one that demands transparency and accountability from frontier developers. So we're requiring in the FRONTIER Act developers to create mandatory safety plans. We're not prescribing what has to be in them, but we're saying that they have to address a certain list of catastrophic risks.

We're requiring them to be transparent about those plans and publish them. And we're requiring accountability because we're requiring them to hire independent verification organizations, not the government, but people that are certified independent of the developers to confirm that they are following their safety plans that they imposed. And we think that that really meets the moment.

BROWN: And we'll talk more about the FRONTIER Act later, but, you know, you have the safety aspect of this and you have the economic incentives, right? "The Washington Post" reports that Trump and his family have a lot of investments in A.I. The decisions the President makes could impact the value of those A.I. investments. How can we trust his administration to make decisions on this that are in the best interest of the public and not just the President, but members of Congress, too, who are able to have investments in A.I. stock?

OBERNOLTE: Sure. Well, I mean, it's an important question, Pamela. And I think the transparency is the answer to that. First of all, we have a balance of powers. So it's not just the administration that's driving decisions about A.I. regulation. It's in concert with the legislative branch, members of Congress, as you point out. And our judicial branch. And that's why we have transparency requirements so -- where -- everyone files a financial disclosure statement to make the public aware when a conflict of interest might occur.

[11:14:59]

And in Congress, we have something called the STOCK Act that restricts our ability to buy and sell the stocks without being transparent about it. And, you know, I hope that that would give the public some confidence that folks are making decisions based on public interest and not their own self-interest.

BROWN: But does it concern you, given the economic incentives, even with what part of what you're proposing here, you're leaving it up to the A.I. tech CEOs, in large part, it sounds like. And they all have economic incentives, right? I mean, so much of this economy is propped up on A.I. right now. I mean, how do you balance that with the safety concerns here?

OBERNOLTE: No, I think you used the right word there, balance. And balance is very important. I would push back on the assertion that our legislation leaves it up to the A.I. companies. The whole reason for doing this is to not leave it up to them to establish a unified national framework for A.I. safety. But I think it's also very telling that in recent days, A.I. companies and their CEOs have come together to say, we need some regulation here.

We need a slowdown and a pause and some, you know, thoughtful discourse about the topic of A.I .safety and how we're going to achieve it. You know, to me, I welcome their engagement because it means that we're all on the same team. We're all trying to push for the same thing.

BROWN: Yes, that's fair. I was just talking about the fact that it would require disclosures. And so you're trusting these companies to disclose whether there are any issues, or. And we have seen that more recently with OpenAI. We've seen that with Gemini. They have actually disclosed issues. But the question is, how can you continue to trust them? There's just a lot at stake here. I also want to just get your reaction. Go ahead.

OBERNOLTE: Well, trust but verify, right? The part of our bill, the larger bill we call the Great American A.I. Act, also includes things like whistleblower protections. So it's you trust but verify. You hope that there's mandatory vulnerability reporting that occurs and that the companies follow the rules. But you also create protections for whistleblowers who can come forward and say, hey, I'm very concerned here. My company is not following the rules. And I think that those two things together are what's required.

BROWN: Yes, and we have heard from some of those whistleblowers, right? You had the former anthropic researcher Jacob Coxon come out saying that he's concerned that A.I. could kill us all by the end of the decade. And then you have NVIDIA co-founder and CEO Jensen Huang, who says he disagrees with that. Let's listen to him.

(BEGIN VIDEO CLIP)

JENSEN HUANG, CO-FOUNDER & CEO, NVIDIA: There is zero percent chance that's going to be the end of the world. And so I think that scaring people is unnecessary. It is irresponsible. And I don't know what their motives are.

(END VIDEO CLIP)

BROWN: So who should Americans believe?

OBERNOLTE: Well, this is a fast moving and evolving technology. And so we all need to be skeptical of everything that we hear. The -- I would completely agree that some of the fears about A.I. are overblown. I'll tell you the things that keep me up at night. It's not something out of a "Terminator" movie where an army of evil robots takes over the world. I don't worry about that.

But what I do very much worry about is the use of A.I. in the hands of malicious human actors to do things like steal people's money through cyber theft and cyber fraud. That's already happening. To do things like hack our critical infrastructure. That's already being done by our geopolitical adversaries. And I worry about the use of A.I. to spread misinformation that undermines our democracy and our trust in each other. So taken together, maybe those are equally consequential to an army of evil robots. But I think we all need to worry about those things.

BROWN: And as you noted earlier, you know, you also have the China factor of all this. The President is expected to meet with Xi Jinping, the leader of China, this week. The President keeps saying that the United States needs to win the race against China. But what exactly does winning look like, in your view, when this is just such a fast evolving technology? What does that actually mean?

OBERNOLTE: Right. Well, the United States has a substantial lead over China in both the development and deployment of Frontier level A.I. And China is trying very hard to catch up. President Xi has made a public commitment to surpass the U.S. and become the worldwide leader in A.I. development by the year 2030. And they are devoting vast amounts of resource to making that happen.

And we need to take that part of the competition very seriously. We should take very seriously the fact that two years ago, China passed the U.S., the number of PhDs that's graduating in computer science. And this year, it's likely that they will graduate over double the number of PhDs that we do in computer science. And so, you know, we need to make sure that we're committed, as the President says, to that part of the competition.

BROWN: But do you think that there should be some agreement worked out between the U.S. and China? And you heard Scott Bessent, the Treasury Secretary, indicate that that is something that they're working on. But also, I just want to go back to what is winning against China look like on the A.I. race? Because if you take A.I. to its natural conclusion, that means it's super intelligent, it's smarter than all humans, and it may not be aligned with human values. So help us understand what winning looks like in this race with China.

[11:20:10]

OBERNOLTE: Right. Well, I think you're looking at it in the wrong way. A.I. is already the most powerful tool that mankind has ever invented for the dissemination of human knowledge. It can teach anyone anything they want to learn about in whatever learning style is optimal for them. And that's going to have profound consequences for human empowerment. But A.I. will shortly be the most powerful tool for the enhancement of human productivity that mankind has ever invented.

And that's going to have profound consequences for our geopolitical competitiveness, our economy, and things like the expansion of human longevity and the expansion of prosperity in any country that leads the world in its development. So that's the reason why it's so important to get this balance between protecting people against harms and not over-regulating correct, because we want those benefits, right? And that's why the President says, and I agree with him, that this competition is so important.

BROWN: Yes. And just to be clear, Congressman, we've talked at length about all the benefits A.I. can provide. There's no question about that. But there's also the downside, which I know you're trying to work on with the FRONTIER Act. And so that's why I keep asking to get some clarity when we say winning against China, what that actually means when it comes to A.I., given the upsides and the downsides, if it's taken to its natural conclusion. Does that make sense?

OBERNOLTE: Right. No, it does. But I disagree with you. I don't think we talk about the upsides enough. I mean, we might give them some lip service, but I think we'd spend the majority of time talking about the downsides.

BROWN: Well, to be clear, after Jacob Coxon came out and said that, you know, humanity could end in a decade, I think that certainly got a lot of people paying more attention to what the potential downsides could be and whether it's even worth it to have upsides if we're all going to be extinct in a decade, something that a lot of people also disagree with that, you know, that hypothesis of what could happen, which is important to have these conversations. Congressman Jay Obernolte, thank you so much. Appreciate you coming on the show. I hope you'll come back to continue this conversation.

OBERNOLTE: Of course. Anytime. Thank you.

BROWN: And still ahead, how President Trump is approaching the flurry of dealmaking and diplomacy expected this week at the United Nations. But right now, he's about to meet with New York Mayor Zohran Mamdani again, what will separate this sit down from the last time they got together.

(COMMERCIAL BREAK)

[11:27:11]

BROWN: Well, happening now, President Trump is preparing to go to New York for a big week on the international stage. His visit includes an address to world leaders at the U.N. General assembly on Tuesday. But today he'll meet with New York Mayor Zohran Mamdani and discuss issues affecting the city, according to Mamdani's office. Despite their significant political differences, the two leaders have maintained a surprisingly friendly relationship and have already met twice.

Joining us now is former New York Mayor Bill de Blasio. So, how significant is a presidential visit to New York City during, you know, to meet with the mayor during this time? Tell us about this.

BILL DE BLASIO (D), FORMER NEW YORK MAYOR: Pamela, it's almost never happened. I mean, you're talking about a handful of times in history that a president of the United States has actually gone to Gracie Mansion, you know, the official residence of the mayor, and sat down with the mayor to talk about the needs in New York City. It's quite striking.

And what's fascinating, of course, we're all fascinated by this personal relationship between these two, but I think what's even more powerful is what's motivating Donald Trump to go out of his way to do this. I think he recognizes that Mamdani has been extraordinarily effective with a message of affordability. Really, part of the main reason why the Democrats are talking about affordability all over the country was the example that Mamdani set last year in his election.

And I think Trump really wants to, in a sense, make sure that message doesn't spread even more effectively. So, he's trying to keep Mamdani from turning into even more of a powerful voice against Trump's failures on affordability. I think this is effectively a containment strategy on Trump's part.

BROWN: You think it's containment. Yes, because I was going to ask you what you attribute this cordial relationship between the two men. What's behind this? Do you think there's anything else besides just that?

DE BLASIO: Look, I think Mamdani did something very important in the second visit to the White House. He presented this vision of affordable housing, which obviously is speaking Trump's language, you know, real estate development. And it's a very powerful project. It's actually an idea Sunnyside Yards in Queens that we first presented in my administration, but we couldn't get enough federal buy in.

There's really a chance that something powerful could happen there. It's over 10,000 affordable apartments, amazing project. And I think Mamdani understands that that might actually catch Trump's fancy and get to be the kind of critical mass point where Trump could actually bring in federal investment. So there's a very tangible reason from Mamdani's side.

But for Trump, I mean, I think this is game respects game. I think Trump sees how powerful a communicator Mamdani is. And I really do think, from Trump's perspective, he knows he's losing altitude all the time in terms of approval. He does not -- he -- Trump knows he doesn't have an effective message. And I think he looks at Mandani, in a sense, as the literal opposite, the guy who's kind of mastered the message around affordability.

[11:30:08]

And that message --