US orders Anthropic to block foreign access to advanced AI models, marking first major government intervention in a frontier AI company's operations and signaling stricter control over the sector.

What happened

The Trump administration ordered AI startup Anthropic to disable access to its most advanced AI platforms for all foreign nationals, following discovery that Anthropic's Fable 5 model could be "jailbroken" (security guardrails bypassed). Treasury Secretary Scott Bessent is among officials now negotiating with Anthropic leadership over the specific security concerns.

Why it matters

This is the most significant government intervention into an AI company's operations to date, and Anthropic called the move "disproportionate." The order sets a precedent just as major AI startups—including Anthropic, valued at more than $900 billion（約140兆円）—pursue public listings, and it may deter future model deployments across the entire sector. The move also contradicts a June executive order in which the Trump administration said it would not force compliance on Silicon Valley; officials are now using extraordinary powers to compel AI developers to address security threats.

What to watch

Anthropic warned that such an approach could "halt all new model deployments for all frontier model providers." The controls will likely shape how rival AI companies including OpenAI, Google, and Meta approach security and foreign access going forward, as US frontier AI models are now treated as strategic national assets with access controlled by national security considerations.