Anthropic Calls for Coordinated AI Slowdown If Systems Begin Self-Improving Faster Than Society Can Manage

Anthropic called on leading AI laboratories to commit to a coordinated pause in development if AI systems begin self-improving at rates that exceed society's capacity to manage the risks. The proposal was made public in June 2026 as the AI industry continues a rapid expansion of capabilities and infrastructure.

The call for a potential slowdown is notable coming from one of the most well-funded AI companies in the world. Anthropic has raised billions of dollars and is developing some of the most capable AI systems available. The company's position reflects a view that the risks of unchecked AI development are real enough to warrant pre-agreed limits.

The proposal does not call for an immediate pause. Instead, it asks labs to agree in advance on the conditions that would trigger one, specifically if AI systems demonstrate the ability to improve themselves faster than humans can evaluate and control the process.

The announcement came the same week that Alphabet launched an $80 billion equity offering to fund AI infrastructure, and SpaceX filed for a $75 billion IPO partly aimed at building space-based AI data centers. The scale of capital flowing into AI development underscores the difficulty of any voluntary slowdown.

Anthropic also joined OpenAI and Microsoft in urging Congress to mandate biosecurity screening for synthetic DNA and RNA orders, arguing that AI tools could otherwise be used to assist in creating biological threats.

The company's Claude Mythos model has gained traction in the Japanese government and financial sectors, prompting new cybersecurity working groups focused on AI-assisted vulnerabilities.