Anthropic says U.S. export-control directive forced Fable 5 and Mythos 5 offline for customers

Anthropic says a U.S. government export-control directive has turned access to Claude Fable 5 and Claude Mythos 5 into a compliance problem: the company said the directive covered foreign-national access to both models and that the practical result was disabling them for all customers. That matters because Fable 5 had just been introduced as the public-facing version of Anthropic’s Mythos-class capability, while Mythos 5 remained limited to vetted cybersecurity and critical-infrastructure users.

The strongest evidence is Anthropic’s own extracted statement. According to Anthropic’s statement, the U.S. government issued an export-control directive citing national-security authorities, suspended access to Fable 5 and Mythos 5 by foreign nationals inside or outside the United States, and included foreign-national Anthropic employees in that restriction. Anthropic said access to all other Anthropic models would not be affected.

CNBC, NBC News and TechCrunch each corroborate the broad customer impact: Anthropic disabled Fable 5 and Mythos 5 in response to a government directive. The evidence supports saying Anthropic disabled the models for all customers to ensure compliance with a foreign-national access restriction; it does not prove that the directive itself legally required worldwide disablement or that Anthropic lacked a narrower technical control.

The regulatory attribution is important but still less firmly sourced than the disablement itself. NBC News reports that the letter came from Commerce Secretary Howard Lutnick and involved Commerce Department Bureau of Industry and Security officials, citing an administration official. Axios similarly reports that Lutnick sent a letter to Anthropic CEO Dario Amodei making Mythos 5 and Fable 5 subject to export controls outside the United States and for foreign persons inside the country. No Commerce Department, BIS, Federal Register or directive text is included in the research pack.

Launch coverage explains why the interruption is more than a routine model-access tweak. The Hacker News reports that Anthropic released Claude Fable 5 on June 9, 2026, and described Fable 5 as public or generally available while Mythos 5 stayed limited to vetted cyber defenders and critical infrastructure operators. MacRumors and TechCrunch launch coverage also described Fable 5 as a Mythos-class model intended for broader use, with TechCrunch reporting that some requests could defer to Opus 4.8 and that Anthropic said early data showed at least 95% of Fable sessions used Fable’s own responses.

The date picture remains messy. The assignment frames the shutdown as June 13, CNBC and other media extractions refer to Friday or June 12, and the extracted Anthropic page is dated June 11. The safest current framing is that Anthropic’s access suspension emerged in the June 11-13 window, with disablement corroborated but exact timing not pinned down by an extracted status page, API catalog change or customer notice.

For AI customers and frontier-model competitors, the episode shifts the access question from pricing, safeguards and customer policy toward state control over model distribution. If the directive is as Anthropic and the media reports describe it, eligibility may depend on export-control compliance and foreign-person restrictions, including employee access, not just whether a user has the right subscription or has passed a vendor’s safety gate.

The next evidence threshold is official government documentation. Until a Commerce/BIS order, Federal Register notice, licensing requirement, or named government statement is available, the story is best read as a company-confirmed model disablement attributed by Anthropic to a U.S. export-control directive, with media-supported but not government-source-confirmed reporting on the specific Commerce/BIS chain of authority.