The Commerce Department pulled Anthropic’s two newest AI models from global access 72 hours after launch, citing a jailbreak that Dario Amodei’s team says already exists in OpenAI’s GPT-5.5.
Secretary Howard Lutnick’s letter to Amodei placed Fable 5 and Mythos 5 under export controls barring access by foreign nationals, including Anthropic’s own foreign-born employees. Unable to verify user nationality at scale, Anthropic disabled both models for everyone worldwide. Older Claude models weren’t affected.
The jailbreak involved prompting the model to read a specific codebase and flag software flaws. The government provided only verbal evidence of what it called a “narrow, non-universal jailbreak.” Anthropic found identical capabilities already in GPT-5.5, routinely used by cybersecurity professionals for defensive work. The company warned this standard “would essentially halt all new model deployments for all frontier model providers.”
The “deemed export” clause is the part that should worry every founder. It applies to foreign nationals inside the US — meaning a startup’s H-1B engineering team could be locked out of their own AI stack on a Friday afternoon, over a verbal tip, with zero statutory process. Other reports indicate Amazon flagged the security concerns to the government; Amazon didn’t respond to comment requests.
Chris McGuire at the Council on Foreign Relations called the across-the-board restriction “highly questionable” and the deemed-export provisions “just absurd.” Box CEO Aaron Levie described it as “a big turning point for AI regulation.” A government willing to pull commercial models on a narrow finding sets a precedent with no obvious limit.
Anthropic says the situation stems from a misunderstanding and is working to restore access. No timeline yet.
Nathan Zakhary