Anthropic Calls for Global AI Pause as Self-Improving Systems Outpace Safeguards

Anthropic is urging the world AI labs to slow down. In a report published Thursday, the company -- now valued at $965 billion after a $65 billion Series H round last month -- called for a global pause on frontier AI development, arguing that the latest models are beginning to show signs they could eventually escape human control.

Co‑founder Jack Clark framed it bluntly to the:¹ You want the option to be able to take your foot off the gas and put your foot on the brake. Right now, it is like the AI industry has a gas pedal, but it does not have a brake pedal.

The proposal, detailed in the Anthropic Institute report and covered by AFP/France 24, is not a unilateral stop. The proposal requires multiple major AI companies in multiple countries -- most notably the US and China -- agreeing to stop at the same time, under verifiable rules.

We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology, the report states. The Wall Street Journal notes the company compared the problem to nuclear arms control.

The pushback was immediate. OpenAI CEO Sam Altman, quoted by the New York Post, characterized the call as strategic marketing: It is clearly incredible marketing to say, We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million.

Wharton professor Ethan Mollick offered a more measured take, per the Post: There is a bit of navel‑gazing, some marketing, and a lot of very sincere beliefs about what Anthropic thinks is likely in the near future of AI. The timing is hard to ignore: Anthropic just dethroned OpenAI as the most valuable AI company, and its restricted Mythos model serves as Exhibit A. Critics note the company is simultaneously preparing for an IPO while telling everyone else to slow down.

Anthropic argument rests on data showing AI is already dramatically speeding up AI development itself. The report and the ¹ highlight: 80% of Anthropic code is now AI‑written; reliable task length is doubling every four months; Claude Mythos Preview achieved a 52x speedup on ML training code. I am worried for my kids if we as a society do not have a serious conversation, Clark told the BBC.

The proposal faces an uphill battle in Washington. As AFP reports, US officials argue any slowdown risks handing China a decisive strategic edge. President Trump discussed cooperating with China on AI safety during a Beijing visit, and signed an executive order this week allowing 30‑day reviews. But the BBC notes testing remains voluntary. Major AI firms have not committed to pausing.

For developers and builders, the pause debate is not just policy. If a global slowdown were achieved, it would directly affect API availability, model access, and the pace of capability improvements. The more immediate reality is that no pause is happening. The AI race continues at full speed. For builders, that means the capability curve keeps climbing -- and the safety questions remain unresolved.