AI developer warns: “We could soon lose control”

“We believe it would be good for the world to have the option to slow down or pause the development of particularly advanced models,” the AI ​​developer said in a blog post.

The aim would be to adapt social structures during this time and to establish AI in the interests of society and people.

Anthropic admitted that there was a case for such a move agreement of the major developers of artificial intelligence would be necessary. So far there has been no apparent willingness in the industry to do this. Companies are more likely to outdo each other with plans to invest hundreds of billions of dollars in AI infrastructure such as massive data centers.

Not a one-sided break

Also Anthropic boss Dario Amodei At the World Economic Forum in Davos in January, for example, he emphasized that the AI ​​race with China would make it very difficult to slow down development because there could hardly be an enforceable agreement. Anthropic rejects one-sided breaks: This would only change who is in front.

As a concrete measure, the AI ​​company’s research institute now wants to work with others to develop a control system that will give developers the security that all participants adhere to such a moratorium and that no one is secretly gaining an advantage.

Also a tech billionaire Elon Musk Years ago, he advocated a six-month break in AI development in order to prevent risks. It later became known that he was simultaneously pushing ahead with the founding of his own AI company.

AI that evolves itself

According to Anthropic, the trigger for the current push is the rapid progress of AI models, especially in the ability to improve and develop themselves. If they were able to do this, the risk of people losing control over artificial intelligence would increase. The milestone could be reached faster than many institutions are prepared to do, Anthropic warned.

The company that develops the AI ​​software Claude often acts as a warning and, among other things, called for stricter regulation of the industry in order to prevent risks. President Donald Trump’s White House accused Anthropic of saying this would slow down US industry in global competition. Anthropic recently caused a stir with the Claude Mythos Preview software, which found security gaps in various programs that had remained undetected for decades.