US Commerce Secretary Howard Lutnick took action against Anthropic’s latest Mythos and Fable AI models after warning of the unacceptable risk they could be deployed by or diverted to military intelligence users in China, Russia or other countries of concern.
Lutnick noted the risk in a letter sent to Anthropic CEO Dario Amodei on Friday US time, ordering the company to suspend export of the models to destinations worldwide and all foreign nationals, wherever located, according to a copy of the letter seen by Reuters.
Senior Anthropic technical staff are scheduled to meet with officials at the Department of Commerce in Washington early this week, a Trump administration official said.
Anthropic’s technical staff have met with officials virtually every day since the Trump administration contacted the company on Friday US time, a person close to the company told Reuters.
After the Commerce letter, Anthropic said it would disable access to the models globally.
The government told the company it believes there is a method of bypassing, or “jailbreaking,” a safeguard that would prevent Fable 5 from being used in identifying software vulnerabilities, Anthropic said in a blog post on Friday. The bypass found only “minor” security flaws that other publicly available models can also find, the company said.
Neither the US Commerce Department nor Anthropic immediately responded to requests for comment on the Monday meeting.
More than 80 cyber security executives and experts on Sunday signed an open letter to Lutnick and US National Cyber director Sean Cairncross that supported Anthropic’s position.
In the letter, cyber security leaders at major firms, including Nvidia and Adobe, asked the Trump administration to lift the restrictions on Anthropic.
Relations between the Trump administration and Anthropic ruptured earlier this year after Anthropic refused to allow the US military to use its AI models for domestic surveillance and fully autonomous weapons systems, and the government retaliated by putting it on a national security blacklist.
The San Francisco-based AI startup, which has confidentially filed for a US initial public offering, had previously warned about the hacking capabilities of its Mythos model and held it back from wide release.
On June 9, Anthropic rolled out a public version, called Fable 5, which included what it described as cybersecurity safeguards.
Anthropic worked with the government to test Fable 5 before it was released, the person close to the company said, and received its approval to deploy it.=
