The US Commerce Department issued an order on June 12, 2026, requiring Anthropic to stop exporting and providing access to its latest AI models, Mythos 5 and Fable 5, worldwide and to all foreign nationals, citing national security concerns about potential use by Chinese, Russian, or other hostile military intelligence agencies [1, 2, 3, 4, 5]. Anthropic complied the same day by disabling access to these models globally [1, 2, 3, 4, 5].
US officials focused on a vulnerability where Fable 5’s safety features could be circumvented to reveal software flaws that might aid cyberattacks, although Anthropic stated only minor issues were found comparable to other public AI models [1, 2, 3, 4, 5]. Since the order, Anthropic’s technical staff have met daily with government officials, including National Cyber Director Sean Cairncross, to negotiate potential solutions and restore access [1, 2, 3, 4, 5]. An Anthropic spokesperson said both sides “are rapidly seeking solutions” and the company remains committed to collaborating with the government to protect US critical infrastructure and maintain leadership in cyber defenses [2].
This is the first use of the 2018 Export Control Reform Act by the Commerce Department to regulate AI model exports, marking a new approach to controlling emerging technology [2, 3, 4, 5]. Earlier in 2026, tensions grew after Anthropic refused requests to support military use of its AI for domestic surveillance and autonomous weapons, leading to the company’s national security blacklisting [1, 4, 5].
Anthropic had internally flagged the Mythos 5 model’s cyberattack potential during limited testing starting April 2026 and delayed its public release. The Fable 5 model, launched June 9 with tighter safety controls, is based on Mythos technology [1, 2, 6, 4, 5].
More than 80 cybersecurity experts and industry leaders, including Nvidia and Adobe, publicly backed Anthropic and called for removal of the restrictions [2, 3, 4]. However, some legal experts questioned whether export controls apply to AI models accessed remotely via cloud APIs rather than traditional physical exports, making the Commerce Department’s actions legally uncertain [2, 3, 4, 5].
Financial firms Morgan Stanley and Goldman Sachs have restricted Hong Kong employees’ access to Anthropic’s Claude AI, indicating growing scrutiny of US AI technology overseas [6].
Anthropic CEO Dario Amodei and Commerce Secretary Howard Lutnick are scheduled to attend the upcoming G7 summit in France, where the matter may be discussed amid ongoing negotiations [1, 2, 3, 4, 5].
