- Access Order: The U.S. government has ordered foreign-national access limits for Claude Fable 5 and Claude Mythos 5.
- Customer Impact: Anthropic says compliance requires disabling both models broadly while it seeks narrower controls.
- Security Dispute: The company disputes whether verbal jailbreak evidence justifies recalling commercially used models.
- Policy Stakes: The order pushes U.S. export-control logic from AI chips toward deployed model access.
On June 12, Anthropic received a U.S. directive ordering foreign-national access to Fable 5 and Mythos 5 and at 5:21 p.m. ET cut access to all users as a result. The company’s compliance reaches customers beyond the non-U.S. citizens named in the order.
Customers face broader disruption than the legal target suggests because Anthropic will disable both models while it tries to resolve the restriction, as non-U.S. citizens, including foreign-national employees, fall inside the order’s scope. Anthropic kept the rest of its product line outside the shutdown.
Why the Shutdown Reaches All Customers
Claude Fable 5 is Anthropic’s safeguarded Mythos-class model for general use, while Claude Mythos 5 is a less-restricted variant for vetted cyber defenders and key infrastructure providers. The demanded foreign-national access rules are harder to enforce than country-targeted account blocks because they can include employees, contractors, and users inside otherwise permitted organizations.
With that, compliance with the U.S. order restricting access to technology for national-security reasons has also become an availability problem for ordinary customers.
Fable 5 and Mythos 5 run on the same underlying model family. Controls around one safety tier can affect customers who were not part of the suspected government concern, especially when the order attaches to a person’s status rather than only to account location. For teams that already started testing or using the models in production, identity screening demanded from Anthropic potentially becomes a service-risk issue moving forward.
Two practical questions remain unresolved. Officials have not publicly detailed the national-security concern behind the order, and Anthropic is still trying to persuade the government that narrower controls can satisfy the directive without keeping both models offline for a broad user group.
The Jailbreak Dispute Behind the Order
Anthropic’s national-security dispute centers on possible jailbreaks, meaning an attempt to bypass the AI model’s safety rules. Before the shutdown, Fable 5 had launched with safety routing.
Fable 5’s routing and Mythos 5’s looser access make the new conflict with the government harder to treat as a routine support issue. Anthropic’s previous handling and implicit capability PR of the walled and not generally released Claude Mythos Preview also appears to have given the government an argument to build on.
The prior Mythos release tied the model family to thousands of zero-day findings, so the new controversy lands in a wider debate over whether powerful AI systems compress vulnerability research into operational risk.
Export Controls Move From Chips to Model Access
Anthropic´s relationship with the government had already been going sideways. Earlier in 2026, a Pentagon blacklist fight followed Anthropic´s refusal to remove limitations on military AI use, including autonomous weapons and mass surveillance. The debate now shifts from government use of Claude toward general access of deployed frontier models.
Pentagon Chief Information Officer Kirsten Davies backed the national-security priority argument, framing defense interests as more important than revenue cycles or pre-IPO valuation.
We fully support @POTUS and @SecWar in prioritizing national security and the security of our warfighters, DIB partners, critical infrastructure, international partners and allies. Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation.… https://t.co/GU9kkHqiOu
— DoW CIO Kirsten Davies (@DoWCIODavies) June 13, 2026
Anton Leicht, a Carnegie Endowment technology and international affairs fellow, cast the shutdown as a sign of U.S. policy leverage.
“It shows how irrelevant most other countries have become to AI policy. It seems like neither access to foreign markets nor any retaliation options held by any other country factored into the administration’s decision.”
Anton Leicht, Fellow, Technology and International Affairs Program at the Carnegie Endowment for International Peace (via TIME)
If the U.S. now treats frontier-model access like AI chip sales, foreign governments may have less room to influence deployment policy. Commercial AI customers face a related risk: access to a hosted model can depend on national-security decisions that are not fully visible to them.
Anthropic is trying to restore access while the order’s national-security concern remains publicly unspecified. A concrete next gate is whether U.S. officials narrow the directive enough for Anthropic to bring Fable 5 and Mythos 5 back online.
If officials accept narrower controls, the case could become a model for targeted access rules. If they do not, customers will have to treat frontier-model availability as a policy risk as well as a product decision.
