The US government ordered Anthropic to pull the plug on two of its most advanced AI models. Three days after launch.
On June 12, 2026, a federal export control directive mandated the immediate suspension of access to Anthropic’s Fable 5 and Mythos 5 models for all foreign nationals, including the company’s own employees. The reason: a narrow jailbreak vulnerability that officials say could lead to the exploitation of security weaknesses. Anthropic complied, shutting down access across the board.
Three days from launch to lockdown
Fable 5 and Mythos 5 had been released on June 9, 2026. The government intervened 72 hours later, making this one of the fastest regulatory clampdowns on a major AI product in recent memory.
Mythos itself wasn’t exactly new to the scene. Anthropic had first previewed the model in April 2026, positioning it as a cybersecurity tool designed to identify software vulnerabilities in critical infrastructure. Anthropic initially released it to a small group of partners exclusively under Project Glasswing for defensive cybersecurity purposes before expanding its offerings.
The jailbreak issue in question, according to Anthropic, was both narrow in scope and already known to the company. Anthropic publicly pushed back against the directive, arguing that the cited problems were minor and that the government’s reaction could stifle innovation across the entire AI industry.
Anthropic expressed full disagreement with the directive while still complying with it.
The market wasn’t impressed
Investors noticed. Anthropic’s pre-IPO shares declined sharply following the announcement.
Fable 5, which Anthropic has characterized as a “safer” model compared to Mythos, got caught in the same net. The directive’s scope covered all foreign nationals, a remarkably broad category that swept up Anthropic’s own international workforce alongside its customer base.