Anthropic urges coordinated AI pause

What's happened

Anthropic has called for a coordinated global option to slow or temporarily pause frontier AI development, warning that models are accelerating their own improvement and could enable recursive self‑improvement. The proposal has prompted debate with OpenAI and US officials as both firms race to release models and prepare IPOs.

What's behind the headline?

What's driving the proposal

• Anthropic has produced internal data showing its Claude models now write large shares of the company’s code and accelerate development cycles. That narrowing of human roles is making the company treat recursive self‑improvement as a live risk that needs institutional checks.

The political and commercial pressure

• Companies are racing to ship models and to prepare IPOs; that creates a clear incentive to avoid unilateral pauses. Anthropic is therefore pitching a multilateral mechanism that would verify compliance so a single actor cannot gain advantage.

What will happen next

• Governments will be forced to choose between overseeing model releases or letting industry set standards. The White House has already moved toward pre‑release review; expect more regulatory steps and tougher export or access controls.

Consequences for technology and business

• A verified slowdown will slow public releases but not halt internal R&D; firms will redirect effort into gated testing, enterprise products and recruitment competition. If coordination fails, the fastest actor will gain market and military advantage, which will increase geopolitical pressure to relax limits.

Bottom line

• The debate has moved from hypothetical risk to operational policy. This will increase pressure on regulators to build enforceable verification and on labs to prove they can secure models before wide release.

How we got here

Anthropic has published research and blog posts showing that its Claude models are automating more engineering work and speeding model development. The company has proposed a coordinated slowdown to let alignment research and policy catch up while OpenAI has argued governments must set rules. Both firms are advancing new models as they prepare for IPOs.

Our analysis

Anthropic has said in its institute report and blog posts that AI is increasingly automating engineering work and could enable recursive self‑improvement; the company wrote that "we believe it would be good for the world to have the option to slow or temporarily pause frontier AI development" (Anthropic, quoted across Business Insider, AP News, Al Jazeera). Jack Clark and Marina Favaro have argued the pause would "enable societal structures and alignment research to keep up" (AP News). OpenAI has pushed a different route: Sam Altman and Jakub Pachocki said democratic governments "must ultimately determine the rules" and proposed an international coordinating body to reduce catastrophic risk (Business Insider). The Guardian’s Stuart Russell and New York Times commentary have framed the proposals as part of a larger argument that industry warnings are alternating with rapid product pushes; the NYT described the pattern as companies 'cataloguing the grim future' while continuing fast development (New York Times Business). Business Insider and CNBC coverage has noted both firms are racing to list publicly and to lock in enterprise customers for agent and coding platforms, such as Claude Code and Codex, increasing commercial incentives to keep moving (CNBC; Business Insider). Direct quote examples: "more than 80% of the code merged into its codebase is now written by Claude" (Anthropic, reported in Business Insider) and Anthropic's report phrase "slow or temporarily pause frontier AI development" (Anthropic; AP News). These sources diverge on remedy: Anthropic argues for a verified, multilateral slowdown; OpenAI argues for government‑led review and coordination; journalists note both approaches are shaped by competitive and IPO pressures.

Go deeper

• How would a multilateral verification system detect secret development?
• Which governments are likely to lead coordination on AI slowdowns?
• How will a slowdown affect enterprise access to AI tools?

More on these topics

• Anthropic - Artificial intelligence company

  Anthropic PBC is a U.S.-based artificial intelligence startup public-benefit company, founded in 2021. It researches and develops AI to "study their safety properties at the technological frontier" and use this research to deploy safe, reliable models for

• OpenAI - Artificial intelligence company

  OpenAI is an artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc.

• Sam Altman - President of Y Combinator

  Samuel H. Altman is an American entrepreneur, investor, programmer, and blogger. He is the CEO of OpenAI and the former president of Y Combinator.

• University of Toronto - Public research university in the Greater Toronto Area of Ontario, Canada

  The University of Toronto (U of T) is a public research university with three main campuses in the Greater Toronto Area of Ontario, Canada. Based on the grounds that surround Queen's Park in Toronto, it was founded by royal charter in 1827 as King's College, the provincial university of Upper Canada and its first institution of higher learning. Originally controlled by the Church of England, the university assumed its present name in 1850 upon becoming a secular institution. The university's three campuses are known as St. George, Mississauga, and Scarborough. Its large downtown Toronto location, St. George, is the oldest of the three and comprises 11 colleges, each with varying autonomy over financial and institutional affairs and significant differences in character and history. Its suburban campuses in the Toronto district of Scarborough to the east and city of Mississauga to the west were established in the 1960s, and together make up nearly half of the university's undergraduate student body. The University of Toronto is the largest post-secondary institution in Canada, with more than 100,000 students enrolled and over 720,000 alumni. It offers over 700 undergraduate and 200 graduate...

• White House - Official residence and office of the President of the United States

  The White House is the official residence and workplace of the president of the United States. Located at 1600 Pennsylvania Avenue NW in Washington, D.C., it has served as the residence of every U.S. president since John Adams in 1800 when the national...

• Demis Hassabis - British artificial intelligence researcher (born 1976)

  Sir Demis Hassabis (/ˈdɛ.mɪs/ DE-mis /hɑːˈsɑː.bis/ hah-SAH-bees; born Dimitrios Hassapis, Greek: Δημήτριος Χασάπης, 27 July 1976) is a British artificial intelligence (AI) researcher and entrepreneur. He is the chief executive officer and co-founder of Google DeepMind and Isomorphic Labs, and a UK Government AI Adviser. In 2024, Hassabis and John M. Jumper were jointly awarded the Nobel Prize in Chemistry for their AI research contributions to protein structure prediction. Hassabis is a Fellow of the Royal Society and has won awards for his research efforts, including the Breakthrough Prize, the Canada Gairdner International Award and the Lasker Award. He was appointed a CBE in 2017, and knighted in 2024 for his work on AI. He was also listed among the Time 100 most influential people in the world in 2017 and 2025, and was one of the "Architects of AI" collectively chosen as Time's 2025 Person of the Year.

• Dario Amodei - CEO and co-founder of Anthropic, Ph.D. Princeton University 2011

  Dario Amodei (born 1983) is an American artificial intelligence (AI) researcher and entrepreneur. In 2021, he and his sister Daniela Amodei co-founded Anthropic, the company behind the large language model series Claude. Prior to that, he was the vice president of research at OpenAI. In his capacity as Anthropic's CEO, Amodei often writes on the benefits and risks of advanced AI systems. He is a proponent of an "entente" strategy in which a coalition of democratic nations use advanced AI systems in military applications to achieve a decisive advantage over adversaries while sharing the benefits with cooperating nations.