AI firm warns powerful systems may outpace human control, urges international cooperation on safety measures.
Artificial intelligence company Anthropic has called for a global pause or slowdown in the development of the world’s most advanced AI systems, warning that emerging models are beginning to show signs of exceeding human oversight capabilities.
In a new report, the maker of the Claude AI models said a temporary halt to frontier AI development would allow governments, researchers and institutions more time to strengthen safety measures and align the technology with human interests.
The company stressed that any meaningful pause would require cooperation among major AI developers and governments, particularly the United States and China, to prevent competitors from gaining an advantage by continuing development unchecked.
Anthropic compared the challenge to nuclear arms control, arguing that rapid advances in AI could eventually lead to “recursive self-improvement” — a scenario in which AI systems become capable of improving themselves with minimal human involvement.
Co-founder Jack Clark said the AI industry currently lacks an effective mechanism to slow development when necessary, describing the situation as having “a gas pedal, but no brake pedal.”
The proposal is expected to face resistance from technology companies and policymakers who argue that slowing AI development could hand strategic advantages to rivals, particularly in the growing competition between the United States and China.
Anthropic said it plans to engage governments, scientists, advocacy groups and industry leaders in discussions aimed at developing a coordinated framework for managing the risks posed by increasingly powerful AI systems.