Anthropic Seeks to Lift Export Controls on AI Model Amid Tensions with Trump Administration

Anthropic, the AI company behind the Claude series of large language models, has been engaged in prolonged negotiations with the Trump administration to lift export restrictions on its Claude Fable 5 model, according to a report by Wired. The talks, which began in June, have seen a shift in leadership, with co-founder Tom Brown replacing CEO Dario Amodei as the primary point of contact, according to anonymous sources involved in the discussions.

Context of the Export Control Directive

The U.S. government imposed the export controls three days after Fable 5’s release, citing concerns over potential cybersecurity risks. The directive barred non-U.S. nationals from accessing the model, which Anthropic classified as a “Mythos-class” system—designed with the same core technology as its more restricted Claude Mythos Preview model. According to Wired, the White House was alerted by Amazon researchers shortly after Fable 5’s launch about vulnerabilities that could allow jailbreaks to bypass its safeguards. The administration also expressed concerns about Chinese-affiliated actors potentially accessing the model.

Anthropic removed Fable 5 from public access on June 12, initiating talks with the Department of Commerce’s Bureau of Industry and Security (BIS) to lift the restrictions. The company has yet to issue an official statement on the matter, but internal communications suggest the model is now undergoing further scrutiny.

Shift in Negotiations: From Amodei to Brown

The initial negotiations were led by Dario Amodei, who reportedly faced criticism for his communication style. Sources described him as “hard to deal with” and “not being a good listener,” with some noting his tendency to “rant” during discussions. Wired cited an anonymous participant who said Amodei’s demeanor—marked by “a trumpet-like speaking voice” and “arm-waving gestures”—complicated the dialogue. Since Amodei stepped back, talks have reportedly improved, with Tom Brown, another co-founder, taking the lead.

Brown, who holds no formal title beyond “co-founder” on his LinkedIn profile, has a “nervous nerd” speaking style but is described as more “personable” than Amodei. His approach is reportedly complemented by Sarah Heck, Anthropic’s head of public policy, whose “enthusiastic and disciplined” communication style aligns with the role’s requirements. A source familiar with the discussions told Wired that Brown “can actually engage” and is “not being a weirdo like Dario.”

Implications for AI Regulation and Export Controls

The case highlights growing tensions between AI developers and regulators over how to balance innovation with national security. Fable 5’s restrictions reflect broader concerns about frontier AI models, which experts warn could be exploited for malicious purposes if not properly controlled. The BIS has previously imposed similar measures on other advanced AI systems, including those developed by Meta and Google, citing risks related to “unauthorized use” and “adversarial applications.”

Analysts note that the outcome of these talks could set a precedent for how the U.S. government manages AI exports. “This isn’t just about one model,” said Dr. Emily Zhang, a senior fellow at the Center for Security and Emerging Technology. “It’s about defining the boundaries of permissible AI development in a globalized tech landscape.”

What Happens Next?

The White House has not commented on the status of the talks, but sources suggest the BIS is evaluating whether Fable 5 meets the criteria for “general use” under existing export regulations. If approved, Anthropic could reintroduce the model to a broader audience, though additional safeguards may be required. The company has not disclosed details about any proposed modifications to the model’s architecture or access protocols.

For now, the situation underscores the challenges of governing AI in real time. As frontier models become more powerful, regulators face increasing pressure to act swiftly while avoiding overreach. The resolution of this dispute could shape the future of AI policy—and the pace of innovation—in the U.S. and beyond.