US lifts export block on Anthropic's Mythos AI for 100+ vetted partners

The Commerce Department on Friday cleared Anthropic to distribute its Claude Mythos 5 model to more than 100 approved US institutions, including large companies and government agencies, walking back an export-control order imposed two weeks earlier. That original block — which forced a shutdown of Mythos and the weaker Fable 5 — followed warnings from Amazon and others that the models could be jailbroken for malicious use. In a letter to Anthropic compute chief Tom Brown, Commerce Secretary Howard Lutnick said sufficient safeguards now exist to let ‘trusted partners’ named in an annex access the model without a license, and credited two weeks of daily negotiations.

The move signals an emerging regulatory regime in which Washington gates the release of frontier AI models, with labs committing to government-defined protocols and standards in exchange for distribution. It landed the same day OpenAI shipped GPT-5.6 to its own short list of government-approved partners, underscoring that the framework is being assembled in real time and applied across the industry. The trigger for the original clampdown was reportedly concern that Mythos had reached entities too closely tied to China, said to involve a South Korean telecom provider.

The letter says nothing about restoring Fable 5, though people close to the talks expect its release on an unclear timeline. The arrangement leaves most of the world waiting: non-US governments, foreign companies, and consumers still have no clarity on access, and European officials and other allies have voiced frustration at their new dependence on US export decisions.