Anthropic executives are expected to meet senior US Commerce Department officials in Washington on Monday after the government restricted foreign access to the company’s latest AI models. CEO Dario Amodei is expected to attend the meeting with Commerce Secretary Howard Lutnick as both sides discuss national security concerns surrounding Fable 5 and Mythos 5.
Anthropic disabled access to both models for all customers on Friday after receiving a directive that prohibited access by any foreign national. The restriction applies to foreign nationals outside and inside the United States, including some Anthropic employees.
Government Raises Concerns Over Reported Vulnerability
Fable 5 is a public version of Anthropic’s latest model that includes additional safeguards, while Mythos 5 uses different controls and is limited to selected organizations. Both are versions of Claude Mythos, which Anthropic began testing with a small group of organizations, including US government departments, in April.
Anthropic had acknowledged that releasing Fable 5 carried risks because its capabilities exceeded those of any model the company had previously made widely available. Within days, the government said it had learned of a possible jailbreak that could allow users to bypass the model’s intended controls.
The company said it had received only verbal evidence of the alleged vulnerability. Monday’s meeting is expected to include additional documentation about the issue, according to Reuters.
Anthropic explained its response to the order in an official statement, saying it suspended all customer access because it could not comply while continuing to operate the models normally. Its other Claude models remain available.
Restrictions Renew Tensions With US Government
The dispute follows earlier disagreements between Anthropic and the US government over the permitted military uses of its models. Anthropic has sued the Department of Defense over restrictions connected to that disagreement, although company executives recently described another meeting with White House officials as productive.
It remains unclear whether the Commerce Department will allow Anthropic to restore access following Monday’s discussions. The government has not publicly provided full technical details about the alleged jailbreak or the specific risks it believes the models pose.
Technology Leaders Call for Access to Be Restored
Dozens of cybersecurity professionals and technology executives have signed an open letter asking Lutnick to remove the restrictions. The signatories include current or former security employees from Nvidia, Zoom, Mercedes-Benz, Google, and the US government.
They also asked officials to use an open, scientific, and transparent process when assessing future AI risks. The letter warned that restricting access to advanced models could limit the tools available to cybersecurity defenders while overseas competitors continue developing similar capabilities.
Featured image credits: Fortune Brainstorm Tech via Flickr
For more stories like it, click the +Follow button at the top of this page to follow us.