Senior Anthropic employees are meeting with Trump administration officials in Washington, D.C., on Monday to try to resolve the AI company's recent high-profile dispute with the U.S. government, according to a source. According to a statement from Anthropic, the company received an export control order on Friday citing "national security powers" and was required to "suspend access to its latest AI models Fable5 and Mythos5 by any foreign nationals within or outside the United States."
To ensure compliance with the directive, the AI startup removed all customer access to the aforementioned models.
The sudden move marks a new twist in Anthropic's relationship with the government. Relations between the two sides were already tense: earlier this year, conflicts with the Ministry of Defense escalated. In March, the Department of Defense classified Anthropic as a "supply chain risk," barring defense contractors from using the company's technology due to what it said was a threat to U.S. national security.
Defense Minister Hegseth posted on social platforms on Saturday in response to the government's latest directive that "every day that passes" proves that blacklisting Anthropic is "the right decision."
Anthropic has filed a lawsuit against the Trump administration in an effort to revoke this supply chain risk determination, and the lawsuit is still ongoing.
Fable5 and Mythos5 were released just last Tuesday, just days before receiving the government export control directive. Anthropic worked with government agencies to test the models before launch and received approval for deployment, according to a person familiar with the matter who asked not to be named.
The government called Anthropic at 1:00 p.m. ET on Friday and instructed the company to disable Fable5 and Mythos5, citing an unspecified national security threat, the person said. Then around 5:30 p.m., Anthropic received a formal letter asking it to suspend use of the models.
Prior to this, Anthropic stated that it had not received any communication regarding national security threats.
Fable5 and Mythos5 are built on Anthropic's powerful Claude Mythos Preview model, which excels at identifying software security vulnerabilities. Anthropic has only opened access to a select group of companies as part of a cybersecurity initiative called Project Glasswing. That approach appears to have earned it some goodwill from the Trump administration, which has held multiple meetings with the company about the model's capabilities.
Anthropic hailed Fable5 and Mythos5 as top cutting-edge models in multiple industry benchmarks. While Mythos5 is still limited to a select group of users, Anthropic has opened up Fable5 to enterprise customers and paid subscriptions. The company said it was ready for widespread release thanks to added security measures to block answers in specific high-risk areas, including cybersecurity and biology.
Anthropic said in a statement on Friday that it believes the government's concerns center on a "potential, non-common jailbreak vulnerability" that allows users to circumvent cybersecurity protections, requiring Fable5 to "read a specific code base and fix any software flaws."
Anthropic stated in a statement: "We do not agree with recalling a commercial model deployed to hundreds of millions of users simply because a potential, localized jailbreak vulnerability was discovered. If this standard were applied to the entire industry, we believe it would essentially halt the deployment of new models by all leading model providers."
Anthropic characterized the dispute as a "misunderstanding" and said it was working to restore access to the model "as quickly as possible."