The U.S. government is allowing Anthropic to restore access to its advanced cybersecurity AI model, Mythos 5, but only for a limited set of U.S. organizations. Multiple outlets report that after weeks of negotiations, the approval follows an earlier government decision that abruptly restricted Mythos 5 access—particularly to foreign nationals—over concerns that safeguards (“guardrails”) could be circumvented. The latest clearance reportedly comes via a letter from U.S. Commerce Secretary Howard Lutnick to Anthropic, acknowledging that the company’s efforts to address security concerns have made “significant progress.” Outlets describe the permission as a partial re-release or redeployment, enabling Anthropic to deploy Mythos 5 to selected customers and partners, including government-related cybersecurity entities. Several sources also note uncertainty around which specific companies will be chosen, with some reporting only that the group is small and that names have not been publicly confirmed. While Mythos 5 access is being reinstated in a limited form, at least one outlet reports that a related model, Fable 5, is not cleared for re-release at the same time, though discussions about widening access may continue.