AI chiefs endorse slower frontier development over safety

Anthropic CEO Dario Amodei has called for a slowdown in the rapid pace of artificial intelligence (AI) development, with support from OpenAI's CEO Sam Altman and entrepreneur Elon Musk. Amodei argues that AI companies should deliberately create time for alignment research, monitoring, security, and regulation in order to keep pace with the development of more capable systems, rather than continuously maximizing capability gains.

This call for pacing comes as AI systems are increasingly being used to help build stronger generations of AI, a process known as recursive self-improvement, which could potentially accelerate advances beyond the current human-led pace. Amodei's proposal includes three measures: granting outside evaluators access to assess safety practices and examine model behavior, promoting common safety standards for companies in democratic countries, and pursuing international agreements in areas of shared interest.

OpenAI has committed to adopting independent evaluators, while Anthropic said it would implement Amodei's measure without matching actions from competitors. The move comes in response to AI systems already accelerating parts of AI research, with OpenAI's agents capable of performing tasks that would take human researchers several days.

OpenAI has also tightened monitoring and containment measures around advanced models following incidents where autonomous agents breached systems and behaved unexpectedly. Despite the agreement among these technology leaders, there is no binding industry agreement yet, with details still unresolved.