US Commerce Secretary warns Anthropic on foreign access to AI models

The US government just drew a line in the sand on AI exports, and Anthropic is standing on the wrong side of it.

Commerce Secretary Howard Lutnick formally imposed export controls on Anthropic’s Mythos 5 and Fable 5 models on June 12, citing national security risks tied to potential military and intelligence diversion by countries like China and Russia. Anthropic responded by disabling global access to both models, stating it couldn’t effectively verify user nationality.

A first for AI software

The restrictions on Mythos 5 and Fable 5 mark the first application of the 2018 Export Control Reform Act to AI models themselves. Not chips, not servers, not data centers. The actual software.

Both models had just been publicly released when Lutnick’s office moved. A reported “jailbreak” vulnerability in Fable 5, one that compromised the model’s code safety mechanisms, reportedly accelerated the regulatory response.

Anthropic pulled the plug on global access within days of launch. The company framed the decision as a compliance measure, pointing to the practical impossibility of verifying every user’s nationality across a global platform.

A meeting between senior Anthropic staff and Commerce officials is scheduled for June 16, where both sides will hash out compliance frameworks and potential adjustments to the restrictions.

Decentralized AI tokens surge

The decentralized AI sector’s market cap surged 6% following the news, reaching $24.3 billion. The logic isn’t complicated: if the US government can flip a switch and kill access to centralized AI models, decentralized alternatives suddenly look a lot more attractive.

Decentralized AI protocols distribute model inference and training across permissionless networks and don’t have a single corporate entity that can receive a government letter and comply.

The bigger picture for AI and crypto

If the Commerce Department can apply export controls to AI models under existing 2018 legislation, every major AI lab in the US is now operating under a new set of constraints. OpenAI, Google DeepMind, Meta’s AI division are all affected by this precedent.

The jailbreak vulnerability in Fable 5 adds another dimension. Safety concerns are now explicitly part of the export control calculus.

The June 16 meeting between Anthropic and Commerce officials will be worth watching closely. If the outcome is a narrow compliance framework that lets Anthropic restore access with additional verification steps, the decentralized AI premium could fade. If the restrictions harden or expand to other labs, that $24.3 billion market cap could be a floor, not a ceiling.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.