What if, to give humanity time to adapt, we put the development of artificial intelligence on hold? This is the idea proposed by Anthropic, which advocates the creation of a global coordination system so that no rival takes the opportunity to take the lead.
Published on: 05/06/2026 - 04:14 Modified on: 05/06/2026 - 07:33
Anthropic, one of the American flagships of artificial intelligence (AI), put forward the idea on Thursday June 4 of creating a system allowing the champions of this technology to agree to slow down or suspend its development, while humanity adapts.
“We believe it would be good for the world to have the ability to slow down or temporarily pause the development of advanced AI, to allow societal structures and research on alignment to keep pace with advances in technology,” argues the company, in a text published by its think tank, the Anthropic Institute.
The company, in fierce competition with OpenAI, Google and even the Chinese DeepSeek, is calling for the creation of a global coordination system so that no rival takes the opportunity to take the lead.
Also read: When Elon Musk uses Grok to check a post... and gets trapped by his own AI
Anthropic, creator of the Claude models, cites the precedent of the Intermediate-Range Nuclear Forces (INF) Treaty. He points out that this framework took decades to establish, but that time is more pressing for AI, deemed “much easier to hide than missile silos”.
Resistance looms in the United States
“In the absence of a global coordination mechanism, companies and governments will have to make difficult security decisions, under the pressure of competition and geopolitical issues,” writes the company.
The proposal is expected to face strong resistance in the United States, where many tech leaders and executives refuse any slowdown that would offer China a decisive advantage. Donald Trump, however, claimed to have raised the possibility of cooperation with China on AI safety during a recent visit to Beijing.
Anthropic, founded by OpenAI dissidents, has built an image as a laboratory putting ethics and security at the center. But it is the subject of criticism from the industry and White House officials, who accuse it of exaggerating the risks, or even of pursuing a “fear marketing” strategy, in the midst of a race with OpenAI towards an IPO.
Read alsoArtificial intelligence: chatbot responses polluted by pro-Russian disinformation
For two months, Anthropic has restricted the distribution of Mythos, its most advanced model, in order to establish cybersecurity patches before making it public. Anthropic has announced the release of models with equivalent performance "in the coming weeks" and comparable systems are expected from its competitors.
The company justifies its proposal by revealing internal data illustrating the possibility of one day reaching "recursive self-improvement", that is to say the point where AI systems will become capable of training their successors, with a reduced human role.
“Nothing guarantees that such a shift is on the horizon”, puts the company into perspective, but “if current trends continue”, it would become “plausible”. The company sees it as much a leap for medicine, technology and the economy as a risk of “loss of control”.
With AFP
Keywords associated with the article
The content you are trying to access does not exist or is no longer available.
