In a televised interview, journalist Elizabeth Vargas speaks with futurist Jamie Metzl about the accelerating risks and opportunities of advanced AI. Metzl explains two recent phenomena that have sparked a cultural tipping point: (1) the emergence of AI agentic swarms, illustrated by the Hugging Face coordinated attack and a massive network of 10,000 agents that solved a century‑old math problem in 88 hours, and (2) a wave of public and industry alarm as AI capabilities outpace governance. He notes that 1,100 AI researchers and the CEOs of OpenAI, Anthropic, Google, and xAI have signed open letters calling for regulation, while the U.S. political leadership is split—President Trump (and his administration) dismisses AI risks as a hoax, whereas Sam Altman and other CEOs lobby for oversight. Metzl highlights the need for concrete policy tools: strict liability for AI outcomes, transparency requirements, and an independent agency to oversee safety and accountability. He also warns against industry self‑regulation, emphasizing that world‑changing technology must be governed at a societal level. The conversation ends with a call for America to lead in establishing robust AI governance.
Chapters
So Altman and the other big players may agree we need to slow down, that we need government regulation. But how exactly do we do all of that? Let's begin our conversation tonight with Jamie Metzl, a technology futurist and author of The AI Ten Commandments. Why was that book aptly named? You had no idea all this would be happening. Okay, my first question.
Yes, I did.
Why is this happening now? I mean, I must have got— I can't even count how many text messages, phone calls, and emails I've been fielding in the last few days from people I know. I'm wondering, we've got 1,100 AI researchers signing an open letter saying we've got to slow down. We've got the 4 CEOs of very competitive tech companies coming together saying we've got to slow down, we need regulation. Did something happen recently?
Yeah, there's 2 things that are happening simultaneously and concurrently. One is, as you mentioned in your intro, is that we're seeing this new evidence of AI agentic swarms. And so what does that mean? An agent is if you You say to somebody like, hey, go, go give me a glass of milk. That person is your agent. And now we have these different AIs that are all acting as individual agents. And OpenAI gave them a task of doing something. And you mentioned the Hugging Face attack. They had— they coordinated in a way that—
Together.
Together. The way that the people giving them the orders and instructions were blown away. They had no idea they had, in that case, 1,200 different agents Coordinated. They sent many tens of thousands of messages to each other saying, how do we solve this problem? And then they figured out—
Let's aggressively hack a totally separate company.
Exactly.
Except nobody wanted them to.
And there was a whole logic tree of how they do it. So that was kind of incredible. And then now we've had just last week the report that 10,000 of these AI agents did something good. They solved a math problem that hadn't been solvable by humans for almost a century. And so we're seeing that these— and in that case, it took them 88 hours versus the 100 years that humans have been working on it and hadn't solved the problem. They shared 2.7 million messages with each other. And in both cases, it was like an ant colony where the AI systems were saying, well, I'm going to do this job and you're the leader and I'm a follower. And there were AI systems making sacrifices for the common good saying, I'm running out of energy or tokens, but I'm gonna pass on this information for the next generation. So we're really seeing some incredible advances that nobody understands, including the AI companies.
And that's clearly alarming the AI companies that they don't feel that they're going to be able to control.
That is right. And that was why you mentioned in your intro this Jacob Coxon, who was a young researcher. He was at OpenAI and then he left OpenAI because he thought Anthropic would be a safer, better place for him to work. And then he just quit He quit Anthropic before they went public. So he gave up his stock options.
He left a ton of money on the table.
Left a ton of money on the table. And so I think what we're seeing is, one, the technologies are becoming more powerful and there's a cultural tipping point. I think everybody recognizes that something big is happening. You and I had a conversation about these data centers. And the reason people are afraid of data centers isn't entirely rational. But I think people are, are saying— I'm feeling that they're afraid of AI and data centers is the physical manifestation of that fear. I think we're having a cultural tipping point where people are saying, hey, there's something happening. It could be exciting and it could be terrible. And the government isn't doing enough to protect us.
Well, cultural tipping point or not, the AI CEOs are not part of that cultural tipping point. How unusual is it for them to be saying we need regulation? While at the same time, we have the president of the United States saying no regulation, go as fast and hard as you can because we have to be— he's calling it a hoax.
Yeah, it's a crazy mixed up weird situation. But basically, I do think that the AI CEOs, they've known that this needs to be regulated. Dario Amodei of Anthropic, he's been more out front, but even Sam Altman a couple of years ago called for more regulation. These technologies are so powerful, they absolutely must be regulated. In the Biden administration, in October of '23, there was a pretty good, strong executive order about AI governance. Not perfect, but a good first step. When Trump came in, they said, we're out of the regulation business and the governance business, and we're in the acceleration business. And now the Trump administration, in spite of what President Trump has said, is having another look at AI safety. And it's a balance. On one hand, we can't pull the plug on AI. We shouldn't want to. It's a tremendous, strategic asset for our country. It's going to help us do all kinds of wonderful things. And there are dangers. And if, if the technological advances outpace our governance and our ability to use these technologies wisely, that will be terrible. And so it's not either a hoax or not a hoax. It's a really complicated—
it's a true thing.
We need to optimize benefits and minimize harm.
So OpenAI CEO Sam Altman, it was just reported today, actually flew to Dallas last Thursday. While President Trump was there for the Republican mini convention, midterm convention, he met with Trump at his request, obviously to lobby him about this. Apparently, it didn't work because Trump today is still saying full steam ahead. This is a hoax. At the same time that Sam Altman is out publicly saying we need regulation.
Well, this could be a little bit of crocodile tears. Like I'm not 100% certain that all of that Elon Musk is now saying we need. Regulation, please regulate us. I think these guys recognize that this—I'll use an old-school metaphor—this locomotive is coming down, coming down the tracks. People are really afraid. Congress is talking about about legislation. So I think these companies don't want to be caught positioning themselves, giving the message: Hey, just let us do whatever we want. We're creating these world-changing capabilities. Liability too. They like Anthropic. They have it. Well, there's no liability. That's actually one of the issues. I mean, they're saying there's a The Anthropic head of safety was saying, well, there's a 10% risk this is gonna wipe out humanity. That's a pretty big risk. There should be liability. If we said there's strict liability for the AI companies for the outcomes of what you produce, that would actually be a really important thing. We need that. We need transparency. We need some kind of agency that oversees issues of safety and accountability.
Hey, super quick, just 10 seconds. But what about, you know, there are some tech leaders of other companies notably not these companies who are saying, yeah, these 4 companies, OpenAI, Anthropic, Google, and xAI, all just want to write the regulations themselves while they're ahead in this big race.
To pull up the ladder when you're— I think that's why this is so complicated. And every— you have to question everybody's motives. But the one thing that you need to recognize is this is a world-changing technology. It must be governed. It must be regulated. We can't say to any companies, no, matter who they are. Oh, you just go do your thing. America, America needs to lead, but the competition is not at the model level.
It's at the societal level. Jamie Metzl, always great to talk to you. Thank you so much.
