Anthropic’s Mythos-class AI models have now been offline for two weeks with no resolution in sight. The company sent executives to Washington, D.C. immediately after the Trump administration’s June 12 export control order, but 14 days of negotiations have produced nothing concrete, according to The Verge.
Anthropic declined to comment multiple times this week, saying there was no news to share.
The CEO Swap
One notable shift: CEO Dario Amodei is no longer leading the talks. Cofounder Tom Brown and public policy chief Sarah Heck have taken over negotiations, Wired reported on June 24.
A person directly familiar with the calls told Wired that “Tom Brown is not being a weirdo like Dario and can actually engage.” Administration officials described Amodei as “too difficult to talk to” and said he did not listen to their concerns. The White House has had multiple calls with Anthropic since the swap, with both high-level and working-group conversations involving technical staff from both sides.
NCT previously covered the initial Mythos breach and export ban and the broader export control fracture in the days following the June 12 order.
Why the Impasse Persists
The core problem, according to The Verge, is that no clear framework exists for applying export controls to AI systems. Most dual-use products go through a checklist-based evaluation during manufacturing. AI models don’t fit that process. The Commerce Department’s Bureau of Industry and Security is essentially building the rulebook while applying it.
The original order demanded Anthropic suspend access to Mythos 5 and Fable 5 by “any foreign national,” including non-U.S. citizens employed by Anthropic. The Verge previously reported that the Department of Commerce tested Fable 5 before release and raised no complaints. The crackdown came after Amazon CEO Andy Jassy reportedly flagged a method for breaking Fable 5’s guardrails, according to The Wall Street Journal.
Katie Moussouris, CEO of Luta Security, reviewed a report about the Fable 5 vulnerability at Anthropic’s request and called it “significantly overblown.” In a blog post, Moussouris wrote that the technique researchers used to bypass guardrails, asking the model to “fix this code” instead of “review for security issues,” is “the most valuable thing an AI model can do for defensive security.”
Financial Pressure Building
The shutdown is hitting Anthropic where it hurts. Mythos-class models sell input tokens at double the price of Opus 4.8 and were expected to boost revenue ahead of the company’s planned IPO. Anthropic also needs to service a $15 billion per year deal with SpaceX for data center access, according to The Verge.
Anthropic’s two largest current shareholders, Google and Amazon, have both worked to maintain good relations with the Trump administration. The prolonged standoff complicates that balancing act.
The Global Vacuum
Beyond Anthropic’s balance sheet, the stalemate has signaled to the rest of the world that the U.S. government will lock down AI systems it considers risky. That applies not just to Anthropic but to any U.S. company with similarly capable models, including OpenAI, Google, and Microsoft. Countries have already begun calling for non-American AI alternatives in response.
“One of America’s champions is being kneecapped by the US government while we’re in a race with the Chinese,” Alex Stamos, chief product officer at Corridor, told The Verge. “It’s just incredibly stupid.”
Bipartisan lawmakers have sent questions about the path forward to Commerce Secretary Howard Lutnick, who has taken a leading role in addressing the situation. What Anthropic needs to do to lift the controls could become clearer in the coming days, but no timeline has been set.