Anthropic's Mythos 5 returns under selective White House approval

Anthropic's Mythos 5 is back online for a defined set of customers after two weeks of halted access. According to reporting by The Verge and the Financial Times, the Trump administration issued a letter authorizing more than 100 U.S. companies and government agencies to use the model, including access by their non-American employees. TechCrunch confirmed the authorization count.

The negotiation structure remains opaque. Neither the criteria for inclusion nor the duration of the approval have been disclosed. The Financial Times notes that unease persists over Washington's ad hoc regulatory approach, which has now demonstrated it can throttle a frontier model mid-deployment with no published framework.

For Anthropic, the two-week outage represents revenue lost and customer confidence compromised. Mythos 5 is the lab's flagship product. Any organization running production workloads on it now carries a new class of counterparty risk: not model failure, but administrative suspension. That risk does not appear in the published service-level agreements.

The precedent matters more than the resolution. If a model can be selectively turned off by executive action, then every enterprise contract that depends on continuous access must now incorporate that risk into its vendor diligence. The margin compression cycle in frontier models was already pricing in open-weight substitution and inference cost declines. It now also prices in non-technical downtime.

The White House has not published the approval list or the conditions under which access can be revoked. Until that framework exists, Mythos 5 remains a conditionally available asset. That is a new category in the model layer.