US Bars Foreign Nationals From Anthropic's Top AI Models

The United States government has directed Anthropic to cut off foreign nationals from two of its most advanced AI systems, forcing the company to suspend access for users worldwide including employees inside the country. The move underscores the growing friction between efforts to maintain technological leadership and concerns that powerful AI tools could be turned into cyber weapons.

Anthropic received the export-control directive at 5:21 p.m. on June 12 and immediately blocked access to Fable 5 and Mythos 5, according to the company's public statement. The order cited national security but supplied no detailed explanation. All other Anthropic models, including the Claude chatbot, remain available. The restriction applies regardless of a user's location or employment status at the firm.

Mythos 5 is the non-public version previously limited to government agencies and selected corporate partners under the Project Glasswing program. It excels at identifying software vulnerabilities that had gone undetected for years. Fable 5, released publicly on June 9, adapts much of that technology for broader use while disabling its cybersecurity and biotechnology functions. Anthropic stated that it had implemented extensive safeguards against misuse and that it disagrees the existence of a potential narrow jailbreak justifies blocking a commercial model already used by millions.

The company noted that similar code-review capabilities exist in models from other providers, including OpenAI's GPT-5.5. It reviewed an internal report it believes prompted the directive and concluded the concern centered on limited code-correction functions. Anthropic has called for a transparent statutory process for such blocks rather than ad-hoc directives.

No independent confirmation of the specific jailbreak evidence or the exact statutory authority cited in the letter has appeared in other reporting. The number of affected users and the precise technical details shared with officials also remain unverified beyond Anthropic's account.