US Restricts Anthropic's Foreign Access to Top AI Models

The U.S. government, citing national security authorities, issued an export control directive instructing Anthropic to suspend access to its most advanced models, Fable 5 and Mythos 5, for any foreign national, Anthropic said in a company statement (Anthropic blog, Jun 12, 2026). Secondary reporting identifies Commerce Secretary Howard Lutnick as the signatory of the directive letter (Bloomberg; CNBC). Anthropic said it received the directive at 5:21 p.m. ET and, to ensure compliance, 'abruptly disable[d]' the two models for all customers; the company said access to all other Anthropic models 'will not be affected' (Anthropic blog; CNBC).

Anthropic's statement describes the government's technical concern as a jailbreak of Fable 5 specifically. Per the Anthropic blog, 'the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws.' Anthropic says it reviewed a report it believes is the basis of the directive and 'validated that the level of capability displayed there is widely available from other models (including OpenAI's GPT-5.5), and is used every day by the defenders who keep systems safe' (Anthropic blog). Coverage in CNBC and Fortune highlights that Mythos-powered features were touted for advanced cybersecurity capability and were rolled out selectively under Project Glasswing (CNBC; Fortune). The suspension of Mythos 5 is separately precautionary: the government's stated technical concern is the Fable 5 jailbreak finding, not a demonstrated Mythos-specific vulnerability (Anthropic blog; Reuters).

Reporting frames this as a marked escalation in U.S. export-control use, moving restrictions from chips and tooling toward access to finished frontier models (Reuters). Multiple outlets note the action follows an earlier dispute that led to a Pentagon supply-chain blacklist for Anthropic after policy disagreements earlier in 2026 (Reuters; CNN). For practitioners, the episode raises questions about how national-security authorities will assess model risk and which operational thresholds will trigger export restrictions; those are broader regulatory dynamics discussed in coverage rather than claims attributed to Anthropic or the government.

Anthropic is complying with the directive but disagreed publicly that 'the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.' The company also stated: 'If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers' (Anthropic blog). Anthropic said it worked with the U.S. government, the UK AISI, and multiple private third-party organizations to red-team Fable's safeguards for thousands of hours prior to launch and that no tester found a universal jailbreak (Anthropic blog).

Observers will follow whether the government provides a written technical rationale or public guidance clarifying what constitutes a disqualifying jailbreak, since Anthropic says the directive lacked specific written details (Anthropic blog; Reuters). Watch for whether other frontier-model providers receive similar directives, for formal appeals or cooperation channels between providers and regulators, and for any independent red-team reports that either corroborate or dispute the government's jailbreak characterization (Fortune; CNBC).