- US warned Anthropic over foreign access to advanced AI models.
- Government cited export controls; Anthropic suspended its model access.
- Anthropic views directive as disproportionate, citing
US Commerce Secretary Howard Lutnick has warned Anthropic PBC that it cannot offer foreign nationals access to its most advanced AI models without explicit government approval, according to a letter reviewed by Bloomberg News. The communication, sent last week, also flagged criminal and civil penalties if the company does not comply.
The directive targets Anthropic’s Fable 5 and Mythos 5 models and has already led the company to suspend access to both systems while it works through the matter with federal officials.
Why Did The US Government Order Anthropic To Restrict Its AI Models?
The letter, dated Friday, instructed Anthropic not to share its Fable 5 and Mythos 5 models with foreign nationals anywhere in the world unless it first obtains a license from the Commerce Department.
Lutnick did not explain why the restriction was necessary, though the letter referenced existing US laws that permit export controls on civilian technology that could potentially be used by an adversary’s military for intelligence purposes, as Bloomberg reported.
ALSO READ: How Did SpaceX Become More Valuable Than Amazon In Just Days?
Lutnick’s letter, addressed to Anthropic CEO Dario Amodei, stated that the company must apply for an individually validated license before exporting, reexporting or transferring the two models to any location or foreign person, regardless of where they are based.
A Commerce Department spokesperson declined to comment, and Anthropic pointed to its own blog post addressing the controls.
How Has Anthropic Responded To The Government’s Directive?
Following the order, Anthropic disabled access to both models late Friday. Since then, company representatives have held virtual discussions with US officials on specific security concerns, and Anthropic’s technical staff met with Commerce Department officials in person on Monday.
The administration hasn’t publicly explained its reasoning, but Anthropic believes the move followed the discovery that Fable 5, a version of Mythos restricted from cybersecurity tasks, could potentially be “jailbroken” to bypass its safety guardrails.
ALSO READ: Pavel Durov Calls India’s NEET-Linked Ban Ineffective And Misguided, Says 150 Million Users Punished
The order marks one of the most direct interventions by the US government into an AI company’s operations so far, arriving weeks after Anthropic confidentially filed for an IPO at a valuation exceeding $900 billion. Anthropic has called the government’s response disproportionate, saying in its post, “We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.”
The company added that applying this standard industry-wide “would essentially halt all new model deployments for all frontier model providers.” Lutnick, currently traveling with President Trump for the G7 summit in France, said the license requirement stays in effect until further notice.
