The U.S. Commerce Department has ordered Anthropic to suspend the export and transfer of its latest artificial intelligence models, citing concerns that they could be exploited by foreign military intelligence agencies from countries including China and Russia. The directive, delivered in a letter from Commerce Secretary Howard Lutnick to Anthropic CEO Dario Amodei on June 12, requires the company to halt distribution of its Mythos and Fable AI models globally, including to foreign nationals within the United States.

According to sources familiar with the matter, the department is concerned that the Fable 5 model contains vulnerabilities that could be “jailbroken” to assist in identifying software security flaws—capabilities which, if exploited, might pose risks to U.S. national security. While Anthropic acknowledged the existence of minor security issues, the company stated these were comparable to those found in other publicly accessible AI models.

Anthropic, a San Francisco-based startup, began restricting access to its models shortly after receiving the letter on June 12. The move follows growing tensions between the Trump administration and the company, which in early 2026 declined U.S. military requests to use its AI technologies for domestic surveillance and autonomous weapons development, resulting in Anthropic’s placement on a national security blacklist.

The Commerce Department invoked the Export Control Reform Act of 2018 to impose these new restrictions, marking the first time it has used the legislation to regulate emerging AI technologies. Under the order, any export or transfer to a foreign national requires a government license, with noncompliance potentially triggering criminal and civil penalties.

However, some experts question the effectiveness and scope of the enforcement. AI models like those developed by Anthropic are typically accessed remotely rather than physically exported, a channel that existing export controls may not fully govern. This has led to debate over whether the Commerce Department possesses the legal authority to impose such restrictions on AI services delivered via cloud platforms.

Leadership from Anthropic have engaged in ongoing discussions with Commerce officials since the directive was issued. Senior technical staff met with department representatives in Washington on June 15 to negotiate terms for resuming access, with National Cyber Director Sean Cairncross participating in the talks. Secretary Lutnick has maintained regular contact with Anthropic as efforts continue to reach a resolution.

Both Lutnick and Amodei are scheduled to attend the upcoming G7 summit in Evian-les-Bains, France, where further discussions on the issue may take place.

The company had recently launched a public version of its AI model, Fable 5, on June 9, following collaboration with government agents who reportedly approved its cybersecurity safeguards prior to release. Despite this, the Commerce Department’s concerns prompted the swift removal of the models from public access.

Meanwhile, the move has drawn criticism from some cybersecurity leaders. More than 80 industry executives signed an open letter addressed to Lutnick and Cairncross urging the administration to lift the restrictions on Anthropic, arguing the controls may hamper AI innovation and do not effectively address the risks posed.

Neither the Commerce Department nor Anthropic have publicly commented further on the matter following the release of the export control letter. The situation remains fluid as both sides seek common ground on safeguarding emerging AI technologies without stifling their development.