The US government on Friday lifted its block on Anthropic's Claude Mythos 5 AI model, enabling the company to distribute it to more than 100 domestic institutions. The move comes after an export control directive two weeks ago forced Anthropic to disable access to both Claude Mythos 5 and Claude Fable 5 for all customers. The company scored a partial victory in a deal with the administration late Thursday.
The reversal signals a nuanced US approach to regulating frontier AI, balancing national security concerns with the desire to maintain domestic leadership. The earlier block had frozen Anthropic's ability to serve key research and enterprise clients. The government appears to be treating different model capabilities separately in these negotiations.
Sources tell Semafor that talks about Fable 5 remain ongoing, suggesting a slower path to resolution for that model. Only Mythos 5 was cleared under the new agreement. The terms of the deal and the specific reasons for distinguishing between the two models have not been disclosed.
The clearance allows Anthropic to resume serving over 100 US institutions that were cut off two weeks ago. For these customers — likely including universities, labs, and corporate partners — access returns after an abrupt disruption. The partial breakthrough may stabilize Anthropic's domestic business while uncertainty persists for international clients.
Critics argue the deal lacks transparency and could set a precedent for opaque, case-by-case AI regulation without clear public standards. The selective unblocking of Mythos 5 but not Fable 5 raises questions about how the government differentiates AI risk internally.