The Trump administration on June 26 authorized Anthropic to restore Claude Mythos 5 access for roughly 100 US organizations that operate and defend critical infrastructure, ending a two-week standoff that had taken both Mythos 5 and its sibling model Fable 5 offline.
A Two-Week Standoff
Anthropic shut down all customer access to Mythos 5 and Fable 5 on June 12, after the Trump administration issued a directive citing national security concerns over foreign nationals — including some Anthropic staff — accessing the advanced models. The cutoff was immediate and global.
On June 26, Commerce Secretary Howard Lutnick notified Anthropic in writing that “appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.” The authorization covers approximately 100 companies and federal agencies operating in cybersecurity and critical infrastructure. It also reverses the earlier restriction on foreign nationals, explicitly permitting access by non-American employees at authorized organizations.
Mythos Cleared, Fable Still Suspended
The clearance applies only to Mythos 5, the specialized cybersecurity model at the core of Project Glasswing — Anthropic’s initiative to help partners scan critical software for vulnerabilities. Partners in the program have used the model to identify more than 10,000 high- or critical-severity security flaws, including a 17-year-old remote-code-execution vulnerability in FreeBSD.
Fable 5, Anthropic’s flagship general-purpose model, remains suspended. Anthropic said it is “continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.”
A New Governance Precedent
The episode illustrates a new pattern in US AI policy: the executive branch formally controlling whether and how a frontier AI model reaches the market. On the same day Mythos 5 was cleared, OpenAI released GPT-5.6 under a similarly restricted, government-vetted structure, available to only about 20 approved partner organizations via API.