US Pulls the Plug on Anthropic's Top AI Models

Artificial Intelligence & Machine Learning , Next-Generation Technologies & Secure Development US Pulls the Plug on Anthropic's Top AI Models Export-Control Order Forces Shutdown of Fable 5, Mythos 5 Rahul Neel Mani (@rneelmani) , Rashmi Ramesh (rashmiramesh_) • June 13, 2026     Credit Eligible Get Permission Anthropic said the U.S. government did not specify the underlying national security concern in writing. The U.S. government has ordered Anthropic to shut down two of its most powerful artificial intelligence models Fable 5 and Mythos 5 under an export-control directive barring access to foreign nationals. The company has complied, under protest, just three days after their public debut. See Also: Securing AI Workloads With Ubuntu Pro Commerce Secretary Howard Lutnick sent the letter to CEO Dario Amodei on Friday, extending the bar to foreign nationals inside and outside the United States, including Anthropic's own employees, the company said. Since Anthropic could not verify the nationality of every user in real time, it has disabled both models for all customers to ensure compliance. Anthropic said the letter did not specify the underlying national security concern, providing only verbal evidence of the potential jailbreak of Fable 5. Anthropic launched Fable 5 last week as the first publicly available model in its new Mythos class, its most capable tier to date. Mythos 5, the same underlying model with certain safeguards lifted, remained restricted to Project Glasswing, Anthropic's program for vetted cyber defenders and infrastructure providers. Mozilla recently reported resolving hundreds of vulnerabilities using an earlier Mythos model. Both models are priced at $10 per million input tokens and $50 per million output tokens (see: Claude Mythos 5 Can Build Exploits But Can't Power Campaigns). Fable 5 was designed to make Mythos-class capability safe for general release. It uses a layer of classifiers, which are separate AI systems that detect potential misuse before the main model responds. When a query touches cybersecurity, biology and chemistry, or model distillation, Fable 5 routes it to Claude Opus 4.8 instead, and notifies the user. More than 95% of Fable sessions trigger no fallback. Before launch, Anthropic said it worked with the U.S. government, the United Kingdom's AI Safety Institute, private third-party organizations and internal teams to stress-test Fable's safeguards for thousands of hours. No tester found a universal jailbreak - one that could broadly bypass the model's protections across a wide range of tasks, it said. The company said it had created cybersecurity classifiers to stop both the development of specific exploits and larger-scale agentic hacking activities such as reconnaissance, lateral movement and multi-step cyber operations. In testing, Fable 5 prevented all progress on these tasks. Anthropic also introduced a 30-day data retention requirement for all Fable 5 and Mythos 5 traffic to support ongoing jailbreak research and mitigation efforts. The company said the data would not be used for model training and would be deleted after 30 days. Anthropic reviewed the report underlying the government's directive and said the capability it demonstrated is available from other publicly deployed models, including OpenAI's GPT-5.5, which is not subject to similar export controls. "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic wrote. The company said it would publish further technical details within 24 hours. The Trump administration had reportedly tried to get Anthropic to pause releasing Fable 5 before the directive arrived but was unsuccessful, said Axios. The friction came into focus earlier this year, after negotiations between Anthropic and the Department of Defense collapsed, and the DOD designated the AI company a supply chain risk - a move Anthropic has challenged in court (see: Pentagon's Anthropic Fight Draws Rebuke From Ex-DOD Leaders). The export-control order drew sharp reactions across the industry. Dean Ball, an AI policy expert who briefly served in the Trump administration, called the move "simply cartoonish" on X, formerly Twitter. Cybersecurity researcher Peter Girnus said Anthropic had created the conditions for this outcome. "If you describe your product as a munition in every press release, eventually a government takes you at your word," he said. "They wrote the legal predicate themselves and called it a brand." The workforce dimension of the directive is also drawing attention. Gaurav Aggarwal, a technology professional, raised a pointed question on LinkedIn: whether foreign nationals employed at American AI companies are now effectively barred from using the very models they help build, and whether those same models could ultimately be deployed against their countries of citizenship. Anthropic said applying this standard across the industry would "essentially halt all new model deployments for all frontier model providers." The company said the government should have the authority to block unsafe deployments through a statutory process that is transparent, fair and grounded in technical facts, and that this action does not meet those principles. Customers on the Pro, Max, Team, and Enterprise plans who had received complimentary access to Fable 5 through June 22 have also lost access.