AI lab leaders call for slowing frontier AI and stronger rules

In short: Some of the people building the most advanced AI are now publicly asking governments to slow it down and regulate it for safety.

What's going on

In mid-September 2026, several well known AI leaders called for a slowdown in work on “frontier” AI, meaning the most advanced AI models (think of them as the fastest race cars in the AI world). They said progress is moving faster than safety checks can keep up.

Anthropic CEO Dario Amodei argued for an “intentional global slowdown” and for independent auditors to work inside AI companies. An auditor is like an outside inspector who can verify safety claims instead of taking the company’s word for it.

After Amodei’s essay, other leaders, including OpenAI CEO Sam Altman, xAI founder Elon Musk, and Google DeepMind leader Demis Hassabis, publicly agreed that tapping the brakes is needed. They said a key problem is coordination. If one company slows down alone, it can lose out to competitors, so they want rules that apply to everyone at the same time.

What to watch

On the Sept. 18, 2026 episode of the Hard Fork podcast, host Kevin Roose said this unusual unity made him more optimistic. He sees it as a sign that AI safety is becoming a mainstream concern, not just something raised by critics.

There is still no global system that can force a slowdown, and governments disagree on how strict rules should be. The next signal to watch is whether policymakers create enforceable steps, such as required safety testing, independent evaluations, and international agreements.

Source: NYTimes