U.S. allows Anthropic to restore Mythos 5 for cybersecurity users

The U.S. government has allowed Anthropic to restore access to Mythos 5 for a limited group of trusted customers and partners, reversing a federal block that forced the company to take the model offline earlier this month. Commerce Secretary Howard Lutnick told Anthropic that the government was confident in the company’s guardrails, and access will now return to about 100 organizations, including government agencies and private companies.

The restored access is meant for U.S. organizations that operate and defend critical infrastructure. Mythos 5 is Anthropic’s strongest cybersecurity model.

The decision followed a June 12 order from the Commerce Department requiring Anthropic to suspend access to Mythos 5 and Fable 5 for any foreign national. Anthropic received that directive at 5:21 p.m. ET and concluded it had to disable the models for all customers because the order covered foreign nationals inside and outside the United States, including foreign-national employees.

The Commerce Department tied the restriction to national security and pointed to a jailbreak technique. Anthropic countered that the problem appeared narrow and that similar vulnerabilities likely exist across other AI systems.

OpenAI announced that it was limiting rollout of its latest GPT-5.6 family to a small group of government-vetted partners at Washington’s request.

John Coleman of the Foundation for Individual Rights and Expression criticized the process as opaque and too powerful, while OpenAI chief executive Sam Altman did not like the idea of the government picking customers.

Anthropic had already tested a limited-release approach in April, when it withheld Mythos Preview from the public and instead shared it with a select group of more than 50 tech organizations through Project Glasswing. That rollout included more than $100 million in usage credits. Anthropic’s Mythos Preview could find and exploit software vulnerabilities with unprecedented accuracy.