US Restricts Foreign Access to Anthropic Models

On June 12, 2026, the US Department of Commerce issued an export-control directive that, per Anthropic's published statement, suspended access by all foreign nationals to the company's most advanced models, Claude Fable 5 and Claude Mythos 5. Anthropic wrote that the directive applies to foreign nationals "including foreign national Anthropic employees," requiring the company to "abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance." Axios reported that Commerce Secretary Howard Lutnick sent a formal letter to Anthropic CEO Dario Amodei informing the company of the restriction; Reuters and The New York Times reported the action cited national-security concerns. Fox Business reported the administration framed the action as a response to what it described as Anthropic's "recklessness" in the model release process.

Anthropic wrote that the government shared a demonstration of a method of bypassing safeguards - a potential "jailbreak." Anthropic wrote its review found a "small number of previously known, minor vulnerabilities" that "appear relatively simple" and that other publicly available models can identify without requiring a bypass. Reuters quoted Anthropic saying the government provided only "verbal evidence of a potential narrow, non-universal jailbreak." The company described prior red-teaming of Fable 5 involving thousands of hours of testing with government and third-party teams, per Anthropic's statement.

Governments and industry have historically focused export controls on hardware and model-training tools; Reuters and CSIS analysis frame this action as a meaningful escalation because it restricts access to a deployed frontier model rather than hardware alone - the first time a US government directive forced a commercially-deployed, frontier AI model offline globally. Project Syndicate columnist Ren Ito framed the broader policy issue as one of "AI sovereignty," arguing that access options matter more than strict ownership rules. BBC, Washington Post, and Al Jazeera reported diplomatic and alliance tensions following the decision. TheNextWeb reported that Trump told Axios he no longer views Anthropic as a national security threat following a G7 meeting with the CEO, signaling potential easing of restrictions.

Observers should track whether the Commerce Department publishes formal technical findings that clarify the conditions triggering the directive; whether other jurisdictions follow with similar access restrictions or reciprocal measures; how cloud providers, enterprise customers, and regulated industries respond operationally to sudden model access limits; and whether Anthropic or other developers change pre-release red-teaming or disclosure practices. For practitioners, the episode underscores a new operational risk: frontier-model availability can be interrupted by export or national-security actions, with immediate consequences for active deployments.