Senior Anthropic Staff members are meeting with Trump administration officials in Washington, D.C., on Monday to try to resolve the issue. artificial intelligence the latest high-profile dispute between the company and the U.S. government, according to a source close to the company.
The person spoke anonymously to discuss an internal matter.
Anthropic received a export control directive Friday, citing “national security authorities” and ordering the company to suspend access to its latest AI models, Fable 5 and Myth 5″by any foreign national, whether within or without the United States”, according to a declaration.
The AI startup has disabled access to models for all its customers to ensure compliance with the directive.
This unexpected action marks the latest difficulty in Anthropic’s relations with the government, which have been strained after a conflict with the Ministry of Defense increased earlier this year. The DOD has labeled Anthropic a supply chain risk in March, which banned defense contractors from using the company’s technology because it would threaten U.S. national security.
Defense Secretary Pete Hegseth addressed the latest government directive by an article on Saturday, writing that “each day that passes” proves why blacklisting Anthropic was “the right decision.”
Anthropic continued the Trump administration for the purpose of rescinding the supply chain risk designation, and that disputes are in progress.
Read more CNBC tech newsAnthropic revealed Fable 5 and Mythos 5 Tuesday, just days before receiving the export control directive from the government. The company worked with government agencies to test the models before their release and received permission to roll them out, according to a person familiar with the discussions who asked not to be named in order to discuss confidential matters.
The government called Anthropic at 1 p.m. ET on Friday and asked the company to deactivate Fable 5 and Mythos 5 due to an unspecified threat to national security, the person said. Anthropic received an official letter around 5:30 p.m. ET demanding that the company suspend the models.
Before the directive arrived Friday, Anthropic had not received any communications regarding a national security threat, the person said.
Amazon CEO Andy Jassy was among several tech executives who raised concerns with senior Trump administration officials about security risks related to Anthropic’s latest AI models, a White House official confirmed to CNBC.
“As a leading cloud provider serving a large number of private and public sector customers, it is not uncommon for governments to seek advice on potential security risks,” an Amazon spokesperson said in a statement. “When they arise, we do not share the details of these discussions.”
Amazon and Anthropic are deeply linked. Amazon has invested $8 billion in the startup in recent years and pledged in April to spend up to $25 billion in the future, with funding tied to certain business milestones. Anthropic is a major user of Amazon’s Trainium AI chips, as well as its cloud technologies. Amazon also relies on Anthropic’s models for many of its services.
Fable 5 and Mythos 5 build on Anthropic’s Claude Mythos Preview, a powerful model that excels at identifying security vulnerabilities within software. Anthropic limited the deployment to a select group of companies as part of a cybersecurity initiative called Project Glasswing, and its approach appeared to earn it some goodwill from the Trump administration, which held several meetings with the company about the model’s capabilities.
Anthropic touted Fable 5 and Mythos 5 as cutting-edge models that exceed many industry benchmarks. Mythos 5 is still limited to a select group of users, but Anthropic has made Fable 5 available to its enterprise customers and paid subscribers. The company said widespread distribution was possible thanks to new safeguards that block responses in specific high-risk areas, including cybersecurity and biology.
In its Friday statement, Anthropic said it believed the government’s concern was about a “potential narrow, non-universal jailbreak,” in which a user could bypass a cybersecurity guardrail and ask Fable 5 to “read a specific code base and fix any software flaws.”
“We disagree that the discovery of limited jailbreak potential should be a reason to recall a business model deployed to hundreds of millions of people,” Anthropic said. “If this standard were applied to the entire industry, we believe it would end all new model deployments for all frontier model providers.”
Anthropic called the dispute a “misunderstanding” and said it was working to restore access to the models “as soon as possible.”
WATCH: Anthropic Disables New Mythos Class Models Days After Release in Response to Government Directive
