The American company Anthropic proposed on Thursday to temporarily suspend the development of artificial intelligence models to allow humanity time to adapt to the transformations imposed by it. A report by the company specialized in the field, which is behind the “Claude” model, called for the establishment of a global coordination mechanism that ensures everyone’s commitment to this halt, so that no party or competitor takes advantage of the opportunity to get ahead of others.
Published on: 05/06/2026 - 07:16
With indications emerging that the latest models of artificial intelligence could escape human control, Anthropic, a company specializing in the field, proposed on Thursday a temporary halt to the development of the most powerful of these systems.
The San Francisco-based company behind the “Cloud” model said in a report that a global slowdown in the development of advanced artificial intelligence “will likely be a positive thing,” but it warned that if just one company stops, competitors will race to overtake it.
Read also: “Meta” Group begins laying off thousands of employees as part of a restructuring that focuses on artificial intelligence
She explained that achieving an effective moratorium at the same time requires the approval of many major artificial intelligence companies in multiple countries, especially the United States and China, according to rules that everyone can verify.
The company has faced opposition from other parties in the industry and officials in the White House who believe that its focus on the worst scenarios carries an exaggeration of the risks, while accusing it of pursuing a strategy to slow down competitors under the guise of raising fears.
However, the White House has approved Anthropic's "Methos" model, which has not yet been released to the public due to its capabilities in the field of cybersecurity and penetration of the most powerful electronic systems, and is currently only used in a limited number of organizations.
This proposal faces great difficulty in Washington and Silicon Valley, where American officials and company managers believe that any slowdown in the development of artificial intelligence threatens to give China an advantage in progress.
Anthropic compared the problem to nuclear arms control treaties, but said that training AI models in secret is much easier than hiding missiles, and the temptation to continue in secret would be enormous.
The company said it plans to bring together government officials, scientists and competing artificial intelligence companies in the coming months to figure out how to create such a system.
Anthropic added that the call for harmonization comes alongside internal data showing that AI is significantly accelerating the development of AI itself.
Anthropic warned that this acceleration could eventually lead to what researchers call “iterative self-improvement,” whereby an AI system becomes able to teach itself on its own without significant human assistance.
“We are not there yet, and iterative self-improvement is not inevitable,” the report said, but warned that it could happen sooner than most governments and institutions imagine.
Anthropic pointed out that “evidence indicates that the human role is diminishing at every step of the artificial intelligence development process.”
France 24/Reuters
Keywords related to the article
The content you want to browse is no longer in service or is currently unavailable.
