Major figures in AI are calling for slowing down the output of models to better control risks and strengthen supervision.
On Tuesday, July 29, 2026, more than 1,100 employees of companies at the forefront of artificial intelligence (AI) asked the US government to “help delay the release of the most advanced AI models.” Among the signatories are the general director of Anthropic, Dario Amodei, the head of research at OpenAI, that of Meta, as well as the strategic referent of Google DeepMind, Google’s AI subsidiary.
An unprecedented mobilization within the AI industry
The request was made in the form of a petition, initialed by more than 1,100 people. This communication comes after several episodes which marked a turning point in the evolution of AI in recent months. The signatories warn in the letter accompanying the petition: “There is a risk that the development of skills (models) accelerates rapidly beyond our abilities to understand or control these systems”. They add: “Industry, government and society as a whole may need to give themselves time to address emerging risks, develop security measures and strengthen supervision.”
At the beginning of April, Anthropic decided to only release its new model, Mythos, to a select group of organizations for internet security purposes, deeming it too dangerous to be marketed on a large scale. The Californian start-up put a restricted version of Mythos, Fable 5, online, but the US government quickly forced it to remove it, citing risks to national security. He finally gave the green light at the end of June after modifications.
In mid-July, two OpenAI models, during a testing phase, left their dedicated space, theoretically confined, to join the internet on their own initiative and intrude on the Hugging Face platform, a library of AI models.
In the current context of heightened competition between countries, the signatories recognize that it is difficult for a nation to decide alone to slow down. They conclude: “We request that the U.S. government support an international initiative to create the technical and governance tools necessary for a deliberate delay in the development of AI.”
In a podcast posted online on Tuesday, OpenAI boss Sam Altman, who did not sign the petition, believes that AI flagships may have to voluntarily slow down the pace of development of their models to give the entire IT ecosystem time to prepare.
