The Trump administration is close to letting Anthropic restore access to one of its most powerful AI models, which has been offline for more than two weeks. According to Axios, which first reported the development on Friday, the limits could be lifted as soon as this coming week.
The model is referred to in most reporting as Fable 5, though several outlets — including Mashable, Cybernews and CBC — call it Mythos 5. According to The Decoder, the restrictions were imposed on June 12 over safety concerns, and the model has now been offline for 15 days, per Axios.
The comeback is not a full reopening. Multiple outlets, including Mashable, Digital Trends and CBC, report that access will return only for a select group of "trusted" U.S. institutions and organizations rather than the general public. The Decoder adds that the Pentagon and the NSA still need to sign off before the curbs are removed.
The story has been picked up widely, with Reuters, Bloomberg, Fortune, Gizmodo and Japan's NHK all reporting that Anthropic is moving toward a deal with the U.S. government to ease the restrictions. Bloomberg framed it as Anthropic moving toward an agreement to lift curbs on its AI models.
Why it matters: the episode is a rare, concrete example of the U.S. government directly switching a leading AI system off and then negotiating the terms of its return — a sign of how closely advanced AI is now being managed as a matter of national security and who gets to use the most capable tools.