Two weeks after the Trump administration forced Anthropic to shut down its most powerful models, the government has partially reversed course. Commerce Secretary Howard Lutnick, in a letter to Anthropic dated June 26, 2026, authorized access to Claude Mythos 5 for more than 100 selected US companies and government agencies. The Fable 5 model remains unavailable to the public.
Key takeaways
- Mythos 5 unblocked for 100+ US companies and government agencies authorized by the Commerce Secretary
- The unblocking covers non-US employees of those organizations — including Anthropic's own non-American staff
- Fable 5 remains unavailable to the public — the decision does not address this model
- The ban lasted from June 12 to June 26, 2026 — 14 days total
- Anthropic: access restored for organizations "that operate and defend critical infrastructure"
A 14-day ban and its partial reversal
On June 12, 2026, Anthropic took Claude Fable 5 and Mythos 5 offline after the US Department of Commerce issued an order under export control regulations. The underlying concern was that security researchers had bypassed Fable 5's safeguards — a finding the government treated as a national security risk. The ban extended beyond public-facing interfaces to API access and, initially, even Anthropic's own non-US employees.
The reversal is not complete. Howard Lutnick wrote: "I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model" — implying each organization on the list had to meet specific security requirements. The list reportedly covers more than 100 entities: both government agencies and private firms in the critical infrastructure sector.
Fable 5 outside the scope of the decision
Fable 5 is a model with somewhat lower capabilities than Mythos 5, designed for broader access. Anthropic launched it just two days before the ban — on June 10, 2026 — marketing it as a safer alternative to Mythos. When the ban covered both models, the government chose to unblock Mythos 5 first — the one reserved for a narrower group of specialists — leaving Fable 5 on hold.
Anthropic confirmed on X that it "continues to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again." No specific timeline was provided.
Regulatory context: a new AI precedent
The Mythos/Fable case is the first instance in which the US government formally applied export controls to restrict access to an AI model on security grounds — and then partially lifted them. This sets a precedent with far-reaching implications for the industry.
For OpenAI, which announced its own GPT-5.6 family on June 26, 2026 in a similarly restricted mode, the Anthropic case functions as both a warning and a template. OpenAI openly criticized the government approach, arguing it should not become "the long-term default." Yet this very pattern — limited access, gradual expansion — may become the standard for future frontier models.
Why it matters
The partial lifting of the Mythos 5 ban reveals that the Trump administration is building a new governance mechanism for frontier models — not through permanent bans, but through a "trusted partners" system with audited safeguards. This shifts regulatory weight from product regulation to access regulation.
For companies in cybersecurity and critical infrastructure that relied on Mythos 5 for vulnerability analysis, the 14-day interruption was operationally costly. The new model — an authorized entities list — may create barriers to entry that favor large organizations with compliance resources, at the expense of smaller players and independent researchers.
What's next?
- Anthropic is negotiating with the US government to restore Fable 5 for general use — no announced timeline
- The Trump administration is expected to complete its 30-day AI model review by July 2, 2026 — the outcome may define formal rules for the "trusted partners" system
- OpenAI is preparing a full GPT-5.6 release for "the coming weeks" in a similarly restricted access model, which may become the new industry standard
Sources
TechCrunch — Trump Admin releases Anthropic Mythos to be used by more than 100 US companies, agencies
Semafor — US releases powerful Anthropic model Mythos to some US companies
Anthropic on X (June 26, 2026) — Post on X
The Verge — Anthropic's Mythos 5 is back





