U.S. Clears Anthropic to Restore Limited Access to Mythos 5 Cybersecurity Model

A Breakthrough in the National Security Standoff

The United States government has granted Anthropic permission to restore restricted access to its advanced artificial intelligence model, Mythos 5, resolving a high-profile national security dispute. In a letter dated Friday, June 26, 2026, Commerce Secretary Howard Lutnick informed Anthropic’s chief compute officer that the developer had made significant progress in mitigating the risks tied to its covered models. Lutnick stated that the system could now be deployed to certain trusted partners, easing a confrontation that began two weeks prior when the government blocked foreign nationals from accessing both Mythos 5 and its sister model, Fable 5. That mid-June order had prompted Anthropic to completely suspend global access to both systems while launching intensive negotiations with federal officials. Commerce Department spokesman Benno Kass noted that the parties worked diligently over those two weeks to maintain American leadership in AI while protecting national security.

Restoring Project Glasswing and Critical Infrastructure Defenses

Unlike typical consumer-facing applications, Mythos 5 was built specifically for highly technical cybersecurity operations. The model was originally distributed through an exclusive, invite-only program known as Project Glasswing, which served approximately 200 partner organizations. This cohort included prominent technology corporations like Amazon, Microsoft, and Nvidia, which utilized the model to detect software vulnerabilities and bolster digital defenses. Under the newly issued government clearance, Anthropic is moving quickly to redeploy Mythos 5 to a select group of American organizations tasked with defending critical infrastructure. These approved entities span vital sectors, including energy, transportation, healthcare, financial systems, telecommunications, and core software networks.

The Fate of Fable 5 and Future Regulatory Frameworks

While Mythos 5 is returning to these secure environments, it remains unavailable for the general public, and its sibling model, Fable 5, stays offline. Fable 5 represents a more widely accessible version of the same core technology, outfitted with strict guardrails designed to block queries regarding nuclear technology, biological threats, chemical weapons, and cyberweapons. The Commerce Department's recent letter did not lift the restrictions on Fable 5, though Anthropic is currently holding weekend discussions with officials to establish a pathway for its release. Beyond resolving the immediate status of these two models, Anthropic is also collaborating with regulators to develop a long-term policy framework for managing future security reviews. This friction reflects a broader trend of tightening oversight under the Trump administration, which also led competitor OpenAI to restrict the rollout of a new model linked to ChatGPT.

Whether this temporary compromise can serve as a sustainable blueprint remains to be seen, especially as developers and regulators attempt to draft a permanent policy framework while keeping highly capable cybersecurity models out of the public domain.