Anthropic abruptly disabled access to its two most advanced AI models, Fable 5 and Mythos 5, for all users worldwide on June 12, 2026, after receiving an export control directive from the U.S. government citing unspecified national security concerns. The Commerce Department order, delivered at 5:21 p.m. Eastern time, banned any foreign national, including Anthropic employees, from accessing the models inside or outside the United States. Because the company could not reliably verify the citizenship of every user in real time across its cloud platforms, it chose to shut both models down globally for all customers to ensure compliance.
The directive came just three days after Anthropic publicly launched Fable 5, a locked-down version of Mythos 5 that the company had positioned as its safest frontier model to date. The government’s stated concern centered on a potential “jailbreak” method that could bypass Fable 5’s safeguards and allow the model to be used for identifying software vulnerabilities. Anthropic said it had reviewed the demonstration, found only a small number of previously known minor vulnerabilities, and noted that other publicly available models could discover them without requiring a bypass. The company also argued that no tester had yet found a universal jailbreak, and warned that recalling a commercial model deployed to millions of users over a narrow scenario would, if applied industrywide, essentially halt all new deployments of frontier AI.
The shutdown underscores how export controls are shifting from hardware like chips to the AI models themselves as Washington seeks to limit foreign adversaries’ access to advanced capabilities. Mythos 5 had been restricted even before the order because of its ability to detect thousands of software vulnerabilities at scale, leading Anthropic to release it only to about 200 organizations under its Project Glasswing testing program, including the U.S. government. A less powerful public version of Mythos was also pulled under the same directive. The move strained Anthropic’s already tense relationship with the administration, which had previously labeled the company a supply-chain risk after it refused to allow U.S. military use of its models for domestic surveillance and autonomous weapons. President Donald Trump later said he no longer viewed Anthropic as a national security threat and described negotiations with CEO Dario Amodei as “going fine,” but the incident has intensified debate over how regulators should weigh jailbreak risks against the pace of AI innovation.
