Anthropic disabled its Fable 5 and Mythos 5 models worldwide on June 12, 2026, after the U.S. Commerce Department issued an emergency export-control directive barring access by any foreign national. Anthropic received the directive at 5:21pm ET; access to all other Anthropic models was not affected. The government cited a reported jailbreak technique; per Anthropic's blog post, the technique essentially consists of asking the model to read a specific codebase and fix any software flaws. Fortune cites cybersecurity consultant Katie Moussouris describing a simple prompt workflow that converted model outputs into actionable vulnerability tests. Anthropic characterized the jailbreak as narrow and non-universal, said the underlying capability is available in other publicly deployed models including GPT-5.5, and disputes the shutdown scope while complying with the directive. Tom's Hardware reports Trump adviser David Sacks said Anthropic refused to fix the jailbreak before export controls were applied. Business Insider reports a UK founder lost mid-project access and avoided major disruption only because he had a backup plan.
What happened
According to Anthropic's blog post dated June 12, 2026, the U.S. government issued an export-control directive citing national security authorities, suspending all access to Fable 5 and Mythos 5 for any foreign national - whether inside or outside the United States, including foreign-national Anthropic employees. Anthropic stated it received the directive at 5:21pm ET and that the net effect was that it had to "abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance." Anthropic confirmed: "Access to all other Anthropic models will not be affected." The Wall Street Journal and Fortune confirm the directive came from the Commerce Department and name a government letter as the triggering communication.
Architecture note
Fable 5 is built on Mythos, Anthropic's underlying cybersecurity-focused model. Per Anthropic, its safeguards were designed specifically to prevent access to Mythos's advanced cybersecurity capabilities; the 30-day customer data retention policy required for Fable 5 - a policy Anthropic acknowledges carries real costs with customers - was part of its defense-in-depth strategy to detect and mitigate jailbreaks quickly.
The jailbreak and government dispute
Per Anthropic's blog post, the government provided only verbal evidence of a potential narrow, non-universal jailbreak. Anthropic states the technique "essentially consists of asking the model to read a specific codebase and fix any software flaws." Anthropic reviewed a demonstration and found it surfaced previously known, minor vulnerabilities also discoverable by other publicly available models, including OpenAI's GPT-5.5. Fortune cites cybersecurity consultant Katie Moussouris describing the exploit as a simple prompt workflow - summarized by Fortune as "fix this code" - that converted model diagnostic outputs into actionable vulnerability scripts, and quotes her saying defenders need access to Fable for legitimate security work. Anthropic stated: "We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result." Tom's Hardware reports Trump adviser David Sacks publicly said Anthropic refused to fix the jailbreak before controls were applied, and that a Chinese group had reportedly accessed the model.
Anthropic's position
Anthropic is complying with the directive while publicly disagreeing with its scope. Its blog post states that if the standard applied here "was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers." Anthropic also referenced its public policy positions calling for a statutory regulatory process that is "transparent, fair, clear, and grounded in technical facts," characterizing this action as not adhering to those principles.
Practical implications for teams using third-party models
Business Insider reports a UK founder, Sean McDonnell, was mid-project when Fable 5 access was cut, and avoided major disruption only because he had a tested fallback. The episode is a concrete illustration that teams relying on single-vendor frontier-model access face sudden-interruption risk from government actions, export-control scope changes, or policy disputes beyond the vendor's control.
What to watch
Observers should track whether the Commerce Department publishes specifics about the underlying vulnerability or the jailbreak technique; how Anthropic documents mitigations and pursues reinstatement; and whether similar export-control actions are extended to other frontier models or providers. The dispute between Sacks and Anthropic over the severity of the jailbreak and the "refusal to fix" characterization is also likely to surface in Congressional or agency proceedings.
Scoring Rationale #
An emergency U.S. government export-control directive forcing Anthropic to disable two frontier models for all global users is unprecedented - a landmark regulatory intervention affecting hundreds of millions of users and setting a precedent for how national-security authorities can interact with commercial AI deployments. The disputed jailbreak characterization and the Anthropic-vs-Sacks public conflict add ongoing policy significance well beyond the immediate access disruption.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.