Trump Administration Shifts Anthropic Negotiations Away From CEO Dario Amodei

The Trump administration has shifted its direct engagement with Anthropic away from CEO Dario Amodei, choosing instead to negotiate Claude Fable 5 redeployment with cofounder Tom Brown and public policy chief Sarah Heck. According to Wired AI, one official characterized the change by saying Brown is “not being a weirdo like Dario and can actually engage,” citing Amodei’s perceived refusal to address administration concerns about model security vulnerabilities.

Export Controls on Claude Fable 5 Remain in Place

The National Security Agency affirmed on June 12 that guardrails protecting Claude Fable 5 and its restricted variant, the Mythos model, could be circumvented to access more powerful capabilities. This finding prompted the Trump administration to impose export controls the same day, taking the models offline. According to Wired AI, the administration has not yet lifted these restrictions, though multiple rounds of talks with Anthropic have occurred at both high-level and working-group levels involving technical staff from both sides.

Negotiations Focus on Technical Proof Standards

The administration and Anthropic are exploring what technical proof from the company might suffice to alleviate government jailbreak concerns, according to Wired AI. A bipartisan group of lawmakers—representatives Sam Liccardo, Jay Obernolte, C. Scott Franklin, and Ted Lieu—sent Commerce Secretary Howard Lutnick a list of questions about redeployment criteria and timeline, demanding responses by June 26. The Commerce Department’s Bureau of Industry and Security manages the export control decision.

Why Guardrail Verification Complicates Redeployment Talks

According to Wired AI, citing prior reporting by Inner Loop, independent cybersecurity experts increasingly view AI guardrails as temporary safeguards that skilled users and future models will find ways to bypass. This conceptual friction—the gap between guardrails as a compliance stopgap and the government’s need for durable assurance—shapes the difficulty both sides face. If guardrails are inherently circumventable, then the proof Anthropic provides today may lack credibility for long-term export-control decisions.

Why This Matters

The administration’s pivot from Amodei to Brown signals a concrete shift in how the U.S. government will negotiate frontier AI security with major labs: technical credibility and responsiveness now outweigh CEO access. Teams at Anthropic and peer organizations will likely prioritize staffing government working groups with engineers and policy specialists over executive engagement. This precedent—established during Q2 2026 while Commerce Department and National Security Agency officials are still defining redeployment criteria—is likely to persist through the remainder of the year as the administration applies similar pressure to other frontier labs facing export-control reviews.