U.S. orders Anthropic to cut off access to new AI models

Anthropic abruptly cut off access to its newest models after the U.S. government ordered the company to suspend Fable 5 and Mythos 5 for any foreign national, including foreign national Anthropic employees inside the United States. The company said it received the directive at 5:21 p.m. Eastern time on June 12 and had no choice but to disable both models for all customers to comply.

The company said federal officials did not spell out the national-security concern. Anthropic understood, however, that the government believed someone had found a “jailbreak” method that could bypass a safeguard built into Fable 5 to block the model from identifying software vulnerabilities. The result was a broad shutdown of the two models just days after their public launch, a sign that Washington is willing to treat access to frontier AI systems as a security issue, not just a commercial one.

The move deepened a conflict that has been building since February, when Secretary of War Pete Hegseth said on X that he was directing the U.S. Department of War to designate Anthropic a supply-chain risk. Anthropic said that decision followed months of negotiations that stalled over two exceptions it wanted for the lawful use of Claude: mass domestic surveillance of Americans and fully autonomous weapons. In March, Anthropic said it had received a formal letter confirming the designation and planned to challenge it in court.

Anthropic has argued that the government’s actions are unprecedented for an American company, while also maintaining that it wants to keep working with national-security agencies. The company said it had supported American warfighters since June 2024 and was the first frontier AI company to deploy models in the U.S. government’s classified networks. It also said it would provide its models to the Department of War and the national-security community at nominal cost.

The latest restriction lands against a backdrop of Anthropic’s own push for tougher controls on advanced AI. In April 2025, the company submitted recommendations supporting stronger export controls on advanced chips and model weights, arguing that America’s compute advantage needed protection. On June 12, the same day it disclosed the shutdown order, Anthropic also announced a $50 billion investment in American AI infrastructure, with about 800 permanent jobs and 2,400 construction jobs as sites come online throughout 2026. That contrast has sharpened the central question in Washington: whether frontier AI is being fenced off on clear national-security grounds, or selectively pressed in ways that employees at one company see as political pressure.