PBD Podcast: Ex-Anthropic Whistleblower on AI Spies, National Security & Control

SPEAKER_1 (0:01)

When power and intelligence start converging, the stakes stop looking like ordinary tech. On this PBD Podcast, Patrick Bet-David sits down with an ex-Anthropic researcher whose public resignation ignited a huge debate about AI safety, Chinese espionage, and whether the race to build smarter models has already gone too far.

And the core message here is blunt: this is not being treated like a normal business, and that may be the problem. The speaker argues that advanced AI is becoming something closer to a national-security issue than a product cycle. He warns, "Whoever controls AI becomes a god." That's a strong statement. What made him resign? His resignation wasn't framed as a tantrum. He says, "I resigned from Anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is acting responsibly." He stepped away after concluding there was no good plan, and that the industry was racing toward something dangerous. In his view, 2027 could be a turning point. So what's the main concern? What makes this different from typical tech development? A major theme is that AI labs are acting like ordinary startups when they're actually handling "super-weapon technology" that "should probably be nationalized." He compares it to nuclear weapons and argues that no private company should make decisions that could shape unemployment, biological risk, or global instability.

That brings up the China angle. What did he say about that? He says it's "almost certain" that China already has spies inside OpenAI and Anthropic, and adds that the labs have weak background checks for work this sensitive. Interestingly, he argues that if the United States slows down, China may slow too, because both sides are building systems that can be copied, stolen, or distilled. Some might call that fearmongering though, right? He pushes back on that. He says he left in a personal capacity because he was genuinely scared. He also insists there was no financial upside, noting that he walked away from major compensation and wasn't trying to launch a hedge fund or monetize the moment.

What about the companies themselves? Did he draw any distinctions between OpenAI and Anthropic? He describes Anthropic as more mission-driven and aligned, while OpenAI felt more career-driven and chaotic. He also says Dario Amadei is more direct and transparent, while Sam Altman is more business-oriented and harder to pin down.

So what's his vision of the future if we continue on this path?

He believes the danger isn't necessarily an apocalyptic ending, but a world transformed by autonomous systems, robots, and drones. He warns that AI will soon automate research itself, speed up capability gains, and make safety harder to control.

As he puts it, "Their optimistic fast-case is my slow-case," and his fear is simple: "we are moving toward a world where AI is out of our control."