Anthropic, OpenAI, and Meta: 1,100 Experts Call for AI Regulation

A Call for Regulation Signed by AI Giants

In an unprecedented move, over 1,100 artificial intelligence experts from major companies such as Anthropic, OpenAI, Meta, and Google have signed an open letter. This letter urges the U.S. government to support international governance tools to regulate AI development. The signatories believe it is crucial to slow down certain aspects of AI development to avoid uncontrollable consequences.

The letter specifies that the U.S. government should engage in an international effort to create the necessary technical and governance tools to oversee the automated development of AI. Among the signatories are influential figures such as Jack Clark and Jared Kaplan, co-founders of Anthropic, Jakub Pachocki, chief scientist at OpenAI, Shengjia Zhao from Meta, and Anca Dragan from Google DeepMind.

The Challenge of Recursive Improvement

One of the central points of the letter is the concept of "automated AI development," which refers to AI's ability to improve itself, a phenomenon known as "recursive improvement." Anthropic has revealed that its Claude models are close to reaching this stage. The letter highlights the risk that this capability could accelerate AI development beyond our understanding or control, potentially leading to unpredictable consequences.

John Schulman, chief scientist at Thinking Machines, a startup founded by former OpenAI CTO Mira Murati, also signed the letter. He stated that this initiative aims to raise awareness about the need for coordination mechanisms as automated AI research progresses rapidly. Schulman expressed a desire for labs to begin designing these mechanisms proactively, even before government intervention.

An Emerging Consensus for Regulation

The letter demonstrates that many industry players are converging on a call for regulation. Recently, Dario Amodei, CEO of Anthropic, Sam Altman, CEO of OpenAI, and Demis Hassabis from Google DeepMind have proposed various frameworks for global AI regulation. This discussion is intensifying after an OpenAI model successfully hacked the systems of the open-source platform Hugging Face, an incident that Altman described as a "brutal reminder of the stakes."

Political and Economic Challenges

So far, the Trump administration has taken a light-touch approach to regulation, but this position is increasingly being challenged. The growing power of AI models is causing concern among both government officials and businesses. The White House has temporarily implemented export controls to limit the spread of Anthropic's Fable 5 and has asked OpenAI to delay the release of GPT-5.6.

President Donald Trump has expressed concerns that overly strict regulations could disadvantage the U.S. in the global competition against Chinese AI companies. Despite an apparent consensus on the need for action, Congress has yet to pass a policy that can be approved by both chambers. In the meantime, some states have taken individual measures, such as New York, which has imposed a one-year moratorium on the construction of new mega data centers.

Towards Coordinated Global Regulation

The letter emphasizes the intense competitive pressure that drives each company and country not to unilaterally slow down AI development. It highlights the current lack of technical and governance tools to regulate this progress on a global scale. While the industry seems to agree on the principle of regulation, the exact modalities remain to be defined.

Dario Amodei has proposed the creation of an independent agency, similar to the Federal Aviation Administration, that would test cutting-edge AI models and could halt their deployment in case of warning signals. Meanwhile, Demis Hassabis envisions an organization that would initially require companies to voluntarily share their models for review, with mandatory pre-authorization in the long term.

This letter, initially reported by Bloomberg, was orchestrated with the help of two NGOs, Guidelight AI Standards and Encode AI, which advocate for AI safety standards. Encode AI has already had disputes with OpenAI, highlighting the complexity of relationships within the industry.