On the website of Anthropic, which is developing the AI model Claude, published An article calling for slowing or halting the development of artificial intelligence. One of its authors is Anthropic co-founder Jack Clark.
The article states that AI systems have approached the stage of recursive self-improvement – the ability to evolve without human intervention. Anthropics does not know what a world in which artificial intelligence can improve itself would look like.
The article calls on AI leaders to slow the development of the technology to give society time to adapt. However, this requires a global monitoring system that avoids the risks of stealth training of new models.
The world has already created systems to verify other complex technologies – for example, the Intermediate-Range Nuclear Forces Treaty. But creating such systems – infrastructure and trust – took decades. We don’t have much time.
Anthropic was created five years ago by a team at OpenAI led by Dario Amodei. In 2026, Anthropic has achieved success with its Claude Code programming tools. The company’s value exceeded that of its competitor OpenAI (the creator of ChatGPT) and reached $965 billion. On June 1, Anthropic announced its imminent listing on the stock exchange.