Anthropic urges a strategic slowdown in the AI arms race

Anthropic, the AI company behind the assistant Claude, has proposed slowing down the race to build advanced AI systems. The firm argues that a unilateral pause would simply shift leadership rather than solve risks. Instead, it wants time to adapt society and improve safety measures before pushing forward.

The call comes as AI companies pour hundreds of billions into infrastructure, fuelled by fierce global competition. Earlier this year, Anthropic CEO Dario Amodei warned that slowing progress is difficult due to the AI arms race with China. The White House under Donald Trump previously criticised the company for suggestions that might weaken US competitiveness.

Anthropic acknowledges that getting major developers to agree on a pause would be tough. But its research institute plans to work with others on a verification system to enforce any future moratorium. The company also stresses the need for stricter industry regulation.

Meanwhile, Anthropic’s latest software, Claude Mythos Preview, recently uncovered long-hidden security flaws in various programs. This discovery highlights both the risks of rapid AI growth and the potential of careful oversight.

Anthropic’s proposal focuses on balancing progress with safety. The firm wants to use any slowdown to strengthen research and societal preparedness. For now, the debate continues over how to manage AI’s rapid advancement without losing ground to competitors.