US lifts restrictions on Anthropic's Fable 5 and Mythos 5 AI models
Commerce cleared Anthropic’s Fable 5 and Mythos 5 after an 18-day freeze, ending a rare federal clampdown on advanced AI access.

The Commerce Department lifted export controls on Anthropic’s Fable 5 and Mythos 5 models, ending an 18-day standoff that had forced the company to pull both systems offline for foreign nationals, including its own employees.
Anthropic said the department reversed the restrictions on June 30, after first ordering a shutdown on June 12 at 5:21 p.m. ET. The earlier directive cited national security authorities and what the government believed was a jailbreak method for Fable 5, but Anthropic said no tester found a universal jailbreak and that the issue exposed only previously known minor vulnerabilities.

The change gives Mythos 5 an immediate opening with a narrow set of users. On June 26, the administration had already allowed the model to be released to roughly 100 companies and federal agencies, and Commerce Secretary Howard Lutnick said “appropriate safeguards” were in place for trusted partners. That approval did not at first restore access to Fable 5, leaving the newer model under tighter federal limits until the latest reversal.
Anthropic said Fable 5 will again be available to global users on Claude, Claude.AI and Claude Code starting July 2. The company also said it will restore the model on AWS, Google Cloud and Microsoft Foundry as soon as possible. Mythos 5 has already been restored for some U.S. organizations and will continue expanding through Anthropic’s Glasswing cybersecurity program.
The episode has become a test case for how far Washington is willing to go in policing advanced AI deployment. The Commerce action was unusually direct, barring access for any foreign national and applying to Anthropic employees as well as outside users. That scope underscored the government’s view that model access can carry national security risks even before a public breach or known misuse campaign is documented.
Anthropic said it spent thousands of hours before launch on red-teaming with the U.S. government, the UK AISI, private third-party organizations and its own teams. The company also requires 30-day retention of customer data with Fable, which it says helps detect and mitigate jailbreak attempts. For now, the reversal restores two of Anthropic’s most sensitive systems to broader circulation, while leaving in place the idea that federal guardrails can still move quickly when Washington sees a security threat.
This article was produced by Prism’s automated news system from verified source data, official records, and press releases, then run through automated quality and moderation checks before publishing. The system is built and supervised by the people who set the standards it runs under. Read our full AI policy.
Did this article answer your question?


