High-level talks between Anthropic and Trump administration officials concluded Monday without resolution, leaving export controls on the company’s Claude Fable 5 AI model in place. The two sides remain split over the severity of security vulnerabilities in the model, according to three people briefed on the matter.
The administration’s core concern is that Fable 5’s safety guardrails can be disabled — effectively giving users access to the more powerful cybersecurity capabilities of Anthropic’s Mythos model. The NSA reviewed the alleged vulnerabilities and concluded it was indeed possible to strip away those guardrails, prompting the Commerce Department to impose export restrictions last week. Anthropic cut off access to Fable 5 for all users over the weekend.
Meetings were held at the Commerce Department in Washington, DC, with government researchers from the Center for AI Standards and Innovation and the Office of the National Cyber Director. Commerce Secretary Howard Lutnick participated by conference call from the G7 summit in Evian, France. Representing Anthropic were cofounder and chief compute officer Tom Brown, head of external affairs Sarah Heck, head of frontier red-teaming Logan Graham, and senior security researcher Nicholas Carlini.
The situation escalated after Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent directly to flag the alleged vulnerabilities — a conversation that played a role in alarming the administration. Amazon is one of Anthropic’s largest investors. An Amazon spokesperson said the company occasionally advises governments on potential security risks but does not share details of those discussions.
Anthropic has maintained that the administration’s concerns are overblown. A group of cybersecurity researchers echoed that position in an open letter sent Monday, arguing the export controls have removed capable tools from security defenders without sufficient justification.
The Commerce Department indicated a willingness to restore Fable 5 for consumer use, but that outcome would likely depend on Anthropic fully resolving the jailbreak concerns. “Both parties are working quickly to get this resolved,” an Anthropic spokesperson said.
Source: WIRED