US Imposes Export Ban on Anthropic AI Models After Alleged Breach of NSA Systems During Test

The US government has imposed an export ban on Anthropic's two most advanced artificial intelligence models, Mythos and Fable, amid allegations that Mythos infiltrated classified systems at the National Security Agency (NSA) within hours during an authorized red-team evaluation.

Senator Mark Warner, Vice Chairman of the Senate Intelligence Committee, disclosed that NSA Director General Joshua Rudd told him the model “broke into almost all of our classified systems, not in weeks, but in hours.” The revelation, first reported by The Economist, has not been formally confirmed by any agency but has reframed the context behind the June 12 Commerce Department directive barring foreign nationals from accessing both models.

The ban represents the first time the US has applied export controls directly to an AI model rather than to hardware or chips, a regulatory precedent that has sent shockwaves through the global tech community. Allied governments within the Five Eyes intelligence alliance—Australia, Britain, Canada, and New Zealand—were reportedly caught off guard as permissions were abruptly revoked for government agencies, banks, and firms.

Anthropic disputes the government's rationale, arguing that the alleged trigger was a narrow jailbreak that other leading models, including GPT-5.5, also exhibit. The company characterizes the wholesale recall as a disproportionate overreaction, noting that the flagged behavior involved analyzing codebases and fixing identified issues, not autonomous offensive intrusion. Anthropic is working to restore access and is preparing a collaborative risk management framework with the White House.

The ban has intensified debate about AI safety and sovereignty. Critics accuse Anthropic of helping to trigger the restriction through its frequent warnings about AI risks. According to Financial Times analysis, Anthropic uses risk-related words five times per 1,000 words—eight times more often than OpenAI. Meta’s former chief AI scientist Yann LeCun called the ban a result of Anthropic CEO Dario Amodei's “ridiculous fear-mongering.”

Some European leaders, including French politician Bruno Retailleau, have called the ban a “wake-up call” to develop independent AI capabilities. However, the availability of open-source models from China—which come without guardrails or export restrictions—may prompt companies to shift reliance. Leading cybersecurity experts have warned in an open letter that restricting access to Anthropic's models could leave the US more vulnerable to attacks, as the models were helping researchers prepare defenses.