US Export Controls Force Anthropic to Cut All Access to Frontier Models Fable 5 and Mythos 5

In a striking escalation of national-security oversight on advanced AI, the US Commerce Department issued an export control directive on June 12, 2026, ordering Anthropic to immediately suspend access to its newly released Fable 5 and Mythos 5 models for any foreign nationals—including those inside the United States and Anthropic’s own non-US employees. To ensure compliance, the company disabled the models for all customers worldwide, leaving other Claude models unaffected.

The directive, received around 5:21 p.m. ET, cited national security authorities without specifying exact concerns, though Anthropic indicated the government’s focus was a potential jailbreak method that could unlock cybersecurity capabilities. The lab had launched Fable 5 days earlier as a heavily safeguarded “Mythos-class” model after extensive red-teaming with the US government, UK AISI, and third parties, emphasizing defense-in-depth strategies including 30-day data retention.

Mainstream reporting framed the move as part of broader US-China tech tensions, but the order’s scope—applying extraterritorially to foreign persons regardless of location—reveals a maturing pattern of treating frontier AI models as controlled dual-use technologies akin to advanced semiconductors. This builds on prior export regimes and internal US debates over AI risks, marking one of the first instances where a commercial lab was compelled to shutter domestic access entirely to meet foreign-access restrictions.

Credible outlets including Bloomberg, Axios, The New York Times, The Wall Street Journal, CNBC, Wired, and The Guardian independently confirmed the directive, Anthropic’s X statement, and the Commerce Department’s involvement under the Trump administration. The episode underscores accelerating regulatory pressure on AI labs and potential ripple effects for industry innovation, talent mobility, and future model deployments.