Just days after their release, Anthropic’s new frontier AI models, Claude Fable 5 and Claude Mythos 5, have been subjected to an export control directive by the U.S. government, effective June 13, 2026. The directive, citing national security authorities, mandates the suspension of all access to both models for foreign nationals, including those employed by Anthropic.
Anthropic launched Claude Fable 5 and Mythos 5 on June 9, 2026, touting them as “Mythos-class” models representing a significant leap in AI capabilities. Fable 5 was made generally available with built-in safeguards designed to mitigate risks in high-stakes domains such as cybersecurity and biology. In contrast, Mythos 5, which shares the same underlying capabilities but with certain safeguards lifted, was initially intended for a restricted group of trusted partners, including those involved in Project Glasswing and the U.S. government. Anthropic highlighted Mythos 5’s exceptional cybersecurity prowess, noting its performance in exploit development evaluations surpassed that of Claude Opus 4.8.
According to an administration official cited by Axios, the Commerce Department’s swift action stemmed from concerns over potential national security risks after another company reportedly managed to “jailbreak” the Mythos model.
This government intervention comes amidst a protracted and often contentious relationship between Anthropic and various U.S. federal agencies. In July 2025, the U.S. Department of Defense (DOD) awarded Anthropic a two-year prototype agreement worth up to $200 million to develop frontier AI capabilities for national security. However, a significant dispute arose in early 2026 when the DOD reportedly demanded that Anthropic’s contracts include an “any lawful use” clause, which conflicted with Anthropic’s foundational principles against using its AI for mass domestic surveillance or fully autonomous weapons.
Anthropic CEO Dario Amodei publicly refused the DOD’s demand to remove safety guardrails, stating the company “could not in good conscience” grant the request, and that “AI-driven mass surveillance presents serious, novel risks to our fundamental liberties.” In retaliation, Defense Secretary Pete Hegseth designated Anthropic as a “supply chain risk” to national security, a label typically reserved for foreign adversaries, and moved to terminate its contracts. This action resulted in the termination of the $200 million DOD contract. While Anthropic challenged this designation in court, with a federal judge largely blocking it as punitive, the broader policy implications have continued to unfold.
Despite the DOD’s blacklisting, reports from June 2026 indicate that the National Security Agency (NSA) has been actively utilizing Anthropic’s Mythos AI model for offensive cyber operations, with Anthropic engineers reportedly embedded within classified NSA facilities. This deployment is reportedly aimed at infiltrating networks in countries such as China and Iran.
Further highlighting the administration’s stance, the Trump administration issued National Security Presidential Memorandum (NSPM) 11 in June 2026. This memo aims to accelerate AI adoption in national security and directs the termination of contracts with companies that attempt to restrict the government’s use of their AI, though it allows for limited waivers for pressing needs, such as the reported use of Mythos by the NSA. Experts widely interpret NSPM-11 as a direct consequence of the administration’s dispute with Anthropic.
Amidst these tensions, Anthropic has maintained its commitment to responsible AI deployment, emphasizing that powerful technologies carry significant responsibility. The company advocates for AI policy shaped by democratically elected governments to safeguard against AI becoming a tool for surveillance and repression.
In a complex turn, Anthropic also reportedly finalized a classified contract with the NSA and DOD that notably omits the contentious “any lawful use” clause and includes restrictions against using the AI on Americans’ data.