AI policy
Why Anthropic is asking the world to slow down AI — explained
What the warning actually says, and why some people are skeptical of the timing.
On 4 June 2026 Anthropic asked countries to agree on a way to slow frontier AI safely.
One of the leading AI companies just asked the world to consider slowing AI down. That's unusual enough to explain properly — here's the plain-English version.
What they're actually worried about
The concern is something researchers call 'recursive self-improvement' — AI helping to build better AI, in a loop that could speed up beyond human oversight. Anthropic's eye-catching example is its own company: it says more than 80% of the new code in its software is now written by its own AI.
Why it's not just stopping
Anthropic argues that if one company stops while others race ahead, that just hands the lead to whoever is least careful. So it's asking countries and companies to coordinate — agree to slow together, in a way everyone can verify, a bit like arms-control treaties. OpenAI made a similar argument days later.
Sources
- Anthropic warns AI may soon begin recursive self-improvement — Scientific American, 5 June 2026
- Anthropic calls for pause of global AI development — RTÉ, 5 June 2026
- Anthropic AI Safety Warning Meets $35B Compute Deal — Tech Times, 11 June 2026