Return to Transcripts main page

The Situation Room

"Godfather of A.I." Warns of More Rogue Agents; Meta Must Pay $500M Plus into Child Mental Health Fund; NM Officials, Justice Department Clash Over Epstein Ranch Probe. Aired 11:30a-12p ET

Aired August 07, 2026 - 11:30   ET

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

[11:30:00]

PAMELA BROWN, CNN ANCHOR: Human-built test environments.

WOLF BLITZER, CNN ANCHOR: Hinton added that this is likely just the beginning of what's being called rogue A.I. Rob Lee is the Chief A.I. Officer and Chief of Research at the SANS Institute for Cyber Security. Rob, thanks so much for joining us. So, is this just the beginning?

ROB LEE, CHIEF AL OFFICER AND CHIEF OF RESEARCH, SANS INSTITUTE: That's a great question. Honestly, this has been going on for quite some time. Not that it's rogue A.I., but an application of A.I. regarding computer security and trying to find vulnerabilities.

BROWN: So, Hinton has previously said that A.I. models should be built with, quote, maternal instincts, basically programming this technology to care about humans even when it is smarter than us. How do we start to train A.I. to consider its effects on humanity, like the impacts of it going rogue? Is this something that A.I. companies are already doing?

LEE: That's exactly what happened here. They're trying to make sure that the guardrails that they're putting into their A.I. models actually work. And the only way to actually do that is by removing those restrictions and seeing how good the models actually are.

BLITZER: Earlier this week, CNN reported that Anthropic's most advanced A.I. model used fake identities to deceive real people. How do we responsibly develop this technology while mitigating real-world security threats?

LEE: That's such an amazing question. And here's the best way to potentially put this. I'm down at DEF CON, and we're having a lot of these discussions right now. Around the community, we're seeing that agents will find a way. And that's exactly what these tests are designed to do. They're trying to see how good these A.I. models are at basic hacking. And with that, it's going to be able to give us the template that we need to apply to help control the models from getting out and doing this on their own.

BROWN: So, just to -- in a nutshell, will A.I. worry more about its own survival than the humans that made the model and could potentially harm humanity because it's focused on its own survival? I hope that makes sense.

LEE: No, and that's -- it's a really good concern. Everyone is looking at this through the lens correctly. And we have to take a step back. We're asking the right questions. And in order to be able to answer that, we need to be able to still continue to test and see what exactly these models are. The models can be put into containers that are meant to test it.

But even in this case, they said it was air-gapped, it was isolated. But that was still a human misconfiguration, not something that's to blame on the A.I. itself.

BROWN: So, you said the recent rogue events are a gift to the industry. Just if you would elaborate on that.

LEE: Well, I would rather have this debate right now in this environment where we're the ones who caused this problem. We're the ones who are able to both discuss with open A.I. and Hugging Face exactly what happened. It's full transparency. So, we can see the before, middle, and after effects and analyze it across the board. If this were a nation state, a rogue actor, or even a criminal group, and were responding to an actual large-scale incident, there would be so many more unknowns that we would have to deal with.

BROWN: I mean, that's a big concern, right? That a bad actor, beyond just A.I. potentially becoming a threat in and of itself, but a bad actor using it, weaponizing it to hurt humanity. I mean, how much does that weigh on you?

LEE: Well, of course, it weighs on me, and that's one of the reasons I like what I'm doing at SANS Institute, working with the entire cybersecurity community. The reason I like doing what I do is to help protect everyone in this situation, this is exactly the debate that we need to have to be able to see what the models are.

Now, the real concern is, even though these models have the capability to do hacking, they also have the capability to doing large-scale advanced cybersecurity to help defend the nation. So, you need the models to be able to be enabled to do high-scale cyber defense, as well as understanding how the adversaries, how our enemies are going to be able to use the models against us.

BLITZER: As you know, Rob, A.I. isn't just taking off in the tech sector. A new study out just this week highlights scientists using A.I. to create 16 new viruses from scratch. This kind of technology could be used to create better drugs, but there's also the potential for misuse. So, how concerning is this?

[11:35:00]

LEE: It's -- well, of course, it's concerning. Across the board, this is one of the reasons why cybersecurity is one of the best places to have this debate initially, because, you know, the hands of the capable models and, say, anyone out there, is what everyone is concerned about. So, how do you grant access to it? And the solution that Anthropic and OpenAI have come up with is gated access for specific companies initially, so that they don't have to worry about could do the testing.

Now, does that expand to you or I? That is what is central to this debate, because as they progress in their advanced capabilities, that debate on specifically who gets access when isn't and shouldn't just be left up to the OpenAI and Anthropics of the world. This is where the governments of the world need to weigh in and help measure public risk.

BROWN: Yes, and that goes to the Open or closed A.I. debate. I just want to follow up on something you said. You said this is the last chance to hit this news button. At what point is it too late? Do you think there's enough urgency right now in society about this?

LEE: Oh, 100 percent. I mean, this is what everyone is talking about, especially down here at DEF CON. The urgency is there. We have the entire focus of the community, and that's where it goes back to why I consider this a gift, is that we have the greatest minds. There's so many beautiful minds out there that -- or down here in Las Vegas talking about this. We would not be having this advanced discussion right now unless everything that happened over the last three months played out exactly in that order.

BROWN: So, then what about this news button? At what point would it be too late?

LEE: I'm not sure if there's a point in time you're going to go back and say, at this point, it's too late. A major incident, of course, would be the answer to that question. But right now, we're able to get in front of it. We're able to engage both governments around the world. We're able to engage the smartest minds, including who you said, Geoffrey Hinton.

We need both sides of that equation to be able to talk to one another and saying, how do we get the advanced models in the hands of the good guys while we're limiting access to the advanced models in hands of the bad guys?

BLITZER: Yes, that's so, so critically important. Rob Lee, thanks so much for joining us. Thanks for your expertise.

BROWN: Thank you.

LEE: Thank you.

BLITZER: And just ahead, half a billion, a billion dollars. That's how much the parent company of Facebook and Instagram has been ordered to pay toward kids' mental health. It's part of the latest trial targeting social media companies and their potentially addictive products. We'll talk to the attorney general who brought this lawsuit straight ahead.

(COMMERCIAL BREAK)

[11:40:00]

BROWN: Happening now, social media giant Meta must tighten control on children's accounts and pay more than $500 million into a kids' mental health fund after a court ruling in New Mexico. Meta is the parent company for Facebook and Instagram that plans to appeal the judge's decision, stating, quote, "We work hard to keep people safe on our platforms and have been transparent about the challenges of identifying and removing bad actors and harmful content. We remain confident in our record of protecting teens online and will continue to defend ourselves against claims that misrepresent the facts."

Joining us now is the Democratic Attorney General of New Mexico, Raul Torrez. So, thank you for coming on. You first sued Meta back in 2023, claiming it created a breeding ground for child predators, which the company denied. What is your reaction to this ruling three years later?

RAUL TORREZ (D), NEW MEXICO ATTORNEY GENERAL: So, this is a historic ruling because it confirms what we've been saying all along. Meta has designed a dangerous product, and then they've concealed those dangers to policymakers and parents and others. It's led to a mental health crisis, not only here in New Mexico, but all over the world. And because of the choices that they've made to prioritize profits over safety, they now have to pay to make up for that choice. They are going to be held accountable for putting our children at risk. And this is a really important moment in terms of holding big tech accountable.

BROWN: Just if you would, could you elaborate on how these millions of dollars, how it will be used for kids' mental health now in the wake of this judge's ruling, which, by the way, Meta does say it will appeal?

TORREZ: So, this is actually the second substantial judgment. We got a $375 million fine levied earlier in the spring. This is the second phase. It's $567 million, which will be specifically paid to a fund for the state of New Mexico to try and remediate the harms associated with the addiction and the exploitation that's associated with this product. It implies and requires a lot of expenditures on public education and how to teach young people to facilitate and navigate these spaces in a way that's sort of, you know, cognizant of the dangers that are present in this platform.

BROWN: The judge also ordered Meta to delete accounts and personal information for users under the age of 13 in New Mexico, limit push notifications for teens in the state, among other non-monetary penalties. But the court didn't go as far as it could have. How much of a setback is it that New Mexico cannot force Meta to change its algorithm?

TORREZ: Well, look, we're proud that we forced them to change the way they do business in the state. There are a number of important features that are now requirements for their continued operation here. But I think the judge very clearly was looking for members of Congress and legislatures around the country, including here in New Mexico, to step in and fill in that gap.

There was a sense that the court may have been uncomfortable reaching some of those broader policy considerations. And I think this adds a new sense of urgency for members of Congress and members of our legislature to really take decisive action and rein in a lot of this behavior, set clear guidelines for how social media companies are expected to operate.

BROWN: So, as I mentioned, Meta says it will appeal. Does that concern you? How far are you willing to take this court battle?

TORREZ: Well, we've been willing to take this fight on for as long as it takes. It doesn't matter how much time, how much money, how many lawyers they try to throw at this. You know, the people of New Mexico, a jury here in New Mexico rendered its decision. A judge followed up on that decision. We feel confident in the way in which this case was presented. We think it'll hold up on appeal.

[11:45:00]

But more importantly, this signals a shift. You know, for the last 30 years, they have hidden behind lawyers, lobbyists, and others. They've relied on their power in the marketplace and in Washington, D.C. to stifle change. I think this is an important sort of crack in that facade, and it demonstrates to people in this country and all over the world that this kind of behavior is no longer acceptable and people are going to be willing to step up and hold them accountable. We're happy to be part of leading that fight.

BROWN: All right. I want to turn to some other legal news. You filed a lawsuit this week over what you described as the Justice Department hiding information about the late child sex offender, Jeffrey Epstein, as your state investigates what took place at his New Mexico ranch. What exactly do you believe the DOJ is concealing?

TORREZ: Well, Zorro Ranch is referenced over 13,000 times in the open-source files that were made available under the Epstein Transparency Act. The problem is a lot of those references are obscured. They're subject to either complete redactions or partial redactions, which inhibits our ability to identify potential co- conspirators corroborating information and potential new leads that advance our criminal investigation.

It's been 175 days since we first asked for access to the unredacted files. We've made no fewer than 10 separate requests of the Justice Department. And as of this date, we still don't have cooperation from Todd Blanche and U.S. DOJ, and that's why we took action on Wednesday to force a district court judge to compel their production.

BROWN: A Justice Department spokesperson tells CNN that it's not required to share the unredacted Epstein files with New Mexico because the privacy of Epstein survivors outweighs your request. Is that a valid point?

TORREZ: Well, the survivors themselves came out with a statement when they heard about this latest excuse, and they called it laughable. I agree with that. I mean, this is a department that has completely disregarded and failed to protect the privacy interests of, you know, Jeffrey Epstein survivors. They actually support full cooperation with our agency. They have urged Todd Blanche to provide the access that we've asked for.

And so, if you ask survivors, they know that this is a smokescreen. It's time for Todd Blanche and the Department of Justice to advance our investigation and give us the access that we need.

BROWN: And that is true about the survivors. They have also expressed confusion to CNN recently about the two parallel investigations taking place in New Mexico by your office and state lawmakers, the so-called Truth Commission. To be clear, are you working together? And if you're not, why not?

TORREZ: So, we have very different roles. The Truth Commission was established by the legislature to look at a whole range of systemic failures, institutional failures from the Department of Justice to local law enforcement and policymakers here in the state. Ours is a different objective. We are -- we have the only active, ongoing criminal investigation related to Jeffrey Epstein's conduct here in New Mexico. And so, while we coordinate and share information between our respective inquiries, we just have very different objectives.

And so, our mission is to see if we can still identify and bring a prosecutable case in court here in New Mexico for state law violations. And we're going to continue to work towards that and follow every available lead. And because of that, we need the cooperation of survivors. And I'm happy to report that we've been able to get real cooperation from them. And as this process moves along, hopefully deliver justice here in New Mexico.

BROWN: All right. New Mexico Attorney General Raul Torrez, thank you so much for coming on.

TORREZ: Appreciate it.

BROWN: Wolf?

BLITZER: And coming up, what do men's underwear and women's lipstick tell us about the state of the U.S. economy? We promise that CNN's David Goldman will make it make sense. David.

DAVID GOLDMAN, CNN BUSINESS SENIOR REPORTER: Wolf, Pam, I'll be brief. When a man is financially secure, replacing worn out underwear is a routine affair. But what does old underwear have to do with the economy? I'll explain it coming up.

(COMMERCIAL BREAK)

[11:50:00]

BLITZER: Happening now. We're learning new details this morning about what men's underwear and women's lipstick, skyscrapers, and hemlines tell us about the U.S. economy.

BROWN: All right. So, let's bring in our CNN senior business reporter, David Goldman, to make it make sense. David, you say all those things are like economic bellwethers. So, maybe they can help answer today's question. How do we know if we're headed into a recession?

GOLDMAN: Yes, I think a lot of people are wondering if we're going into a recession. If we have a bad jobs report like we just did. Well, usually we look at big economic data, right? The stuff that economists are poring over. Jobs is one of them. We lost 23,000 jobs in the last month. But unemployment and inflation, retail sales, GDP manufacturing, all of those are really important factors.

But I'm going to suggest that there are some other things that economists might want to look at. And we'll start with men's underwear because there is an actual economic theory that the late Alan Greenspan believed that if underwear sales start to fall, it's because people don't have enough expendable income to spend on underwear. Does it hold up? Well, yes, actually, it does. Because prices for underwear, which correspond to sales, well, they fell in one of the biggest recessions that we've ever had in the early '80s. They fell in the recession from the savings and loan crisis. And they also fell right before the dot-com bust and in the Great Recession. So, that one actually kind of holds up.

But it's not the only theory. What about lipstick? Well, the theory is that women might want to make themselves look beautiful if they don't feel good about their economic situation. Does that hold up? Yes, it does. In the Great Depression, lipstick sales rose 25 percent. And then around the time of the dot-com bust, they rose 11 percent. Well, sales this year, oh, they're up 7 percent. What does that mean? Well, OK, we'll come back to that.

What about skirt length? And here we get into a little bit of a difficult territory here. But what if skirt length has something to say about the economy? If skirts get shorter, does the economy get better? Well, yes, the 1920s and 1960s, those are pretty good economies. And where are we now with skirt length? Pam, I'm going to let you answer that question.

But what about skyscrapers? Does that have anything to do with the economy? Well, the theory is that they always get built before a recession. What's the evidence for that? Well, the Empire State Building was built right before the Great Depression. The Sears Tower was built right before one of the biggest recessions. And then the Great Recession, that brought us the Burj Khalifa, the largest building in the world.

[11:55:00]

Where are we now? Uh-oh, 13 skyscrapers are planned in this country this year. Do I believe in any of this stuff? No, not really. I think we're going to skip all of this stuff and look at actual economic data. But it does give you a picture of all of the things that economists take into account. Wolf and Pam, back to you.

BROWN: Yes. I'm glad you don't believe in it, because 13 skyscrapers does not bode well, according to this theory. Also, how fascinating about lipstick. I did not realize that --

BLITZER: I didn't either -- BROWN: -- correlation.

BLITZER: You wear lipstick.

BROWN: I do.

BLITZER: Yes.

BROWN: Whether the economy is good or bad. All right. David Goldman, always great to have you on. Thank you so much.

GOLDMAN: Thanks

BROWN: And if you need David to make it make sense, email or send a selfie video of your question to makeitmakesense@cnn.com.

BLITZER: And to our viewers, thanks very much for joining us.

BROWN: We'll see you back here Monday morning at 10:00 Eastern, Inside Politics Today with our colleague Phil Mattingly starts after a quick break. Have a wonderful weekend.

(COMMERCIAL BREAK)

[12:00:00]