Zuckerberg to AI labs: regulate yourselves — no industry pact needed

What happened#

A few weeks ago, Meta CEO Mark Zuckerberg drew a bright line in the AI safety debate. In a public post, he rejected calls for an industry-wide agreement to slow down AI development when safety worries flare up. His argument: every lab already has both the responsibility and the business incentive to move at a safe pace — no pact required.

"Any lab that doesn't focus on alignment will fall behind," he wrote, meaning AI systems that don't reliably follow users' intentions will lose to ones that do. As proof that self-regulation works, he pointed to Meta's own decision to delay its Muse AI agent by several months to strengthen its security: "We didn't call for everyone else to do this before we would. We just did it."

Not everyone agrees. Anthropic CEO Dario Amodei has called for labs to coordinate slowing down "recursive self-improvement" — AI improving itself — and even asked regulators for antitrust permission so rivals could legally agree to go slower together. The FTC's chairman, Andrew Ferguson, responded that everyone should be "deeply suspicious" of AI companies seeking antitrust exemptions while lobbying for new rules.

Since then, the story has moved to Washington and New York. In late September, leaders from OpenAI, Anthropic, Meta, Google, Nvidia and SpaceX signed a short voluntary accord at the White House promising to self-regulate. And on October 5, AI executives testified under oath at a New York City Council hearing — where all 51 council members introduced 10 bills to oversee AI in the city, including a whistleblower program and a required "kill switch" for AI systems. The message from the council: self-regulation isn't enough.

What it means for you#

This is really a fight about who you trust to keep AI safe: the companies building it, or the governments watching them. Zuckerberg's bet is that the market sorts it out — unsafe AI loses users, so labs will behave. Amodei's bet is that some risks are too big to leave to competition. And cities like New York are betting on neither, writing their own rules instead.

For anyone using AI products, the practical takeaway is simple: the safety standards behind the tools you use are still being negotiated in real time, lab by lab and city by city. There is no single rulebook yet — which is exactly what this whole argument is about.