Anthropic Mythos 5 unblocked for 100-plus US institutions

The US government just demonstrated it can act as a gatekeeper for frontier AI model access, impose restrictions, and lift them within weeks. Semafor reports that the Trump Administration has rescinded export controls on Anthropic's Claude Mythos 5, authorizing the company to distribute the model to over 100 American institutions, including corporations and government agencies.

The block had been in place for roughly two weeks, put there by the same administration citing potential security vulnerabilities. Earlier concerns reportedly centered on Mythos reaching partners with alleged connections to China, specifically a South Korean telecommunications provider. Commerce Secretary Howard Lutnick formalized the reversal in a letter to Anthropic: "I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model." Entities named in the authorization letter no longer require export licenses to use the model.

Not everything is resolved. Fable 5, the sibling model also restricted two weeks ago, still has no clear release timeline, with officials indicating discussions are progressing toward approval but no date confirmed. The broader framework for government oversight of frontier AI models is still developing, leaving international users without clarity on when or whether they might gain access.

The timing is notable: the announcement coincided with OpenAI's release of GPT-5.6 to selected government-approved recipients, suggesting a pattern where leading frontier models are being channeled through a government approval process. What the reporting does not give you is any detail on what specific safeguards Anthropic put in place, or how authorization criteria will be applied consistently going forward. For companies not yet on an approved list, the rules are still being written.