Anthropic’s Export Ban Traced to Amazon Competitor Report, Not Security Data

Published by James Harris on June 20, 2026June 20, 2026

Trump told Axios that Anthropic CEO Dario Amodei is no longer a national security concern in his view, two days after meeting him at the G7 summit in France. The underlying export controls, however, remain intact.

The Commerce Department’s June 12 order requiring federal approval before foreign nationals can access Mythos 5 and Fable 5 has not been lifted, and the Pentagon’s March designation barring federal agencies from using Anthropic technology stands. What the Axios interview added was the origin of that June 12 directive: according to Trump, a competitor who is also an Anthropic investor flagged a vulnerability in Mythos to the administration. That description fits Amazon, which has invested $8 billion in Anthropic while running competing models through AWS Bedrock. A major investor delivering a vulnerability report that triggers a national-security export-control action against the company it partially owns is a structurally unusual sequence, and it raises questions about the evidentiary standard the Commerce Department applied before Secretary Howard Lutnick sent the letter at 5:21 PM on a Friday, giving Anthropic 90 minutes to act.

Anthropic ultimately disabled both models for all users rather than attempt nationality-based access controls it concluded it could not reliably enforce.

The timing matters because the export restrictions on Fable 5 and Mythos 5 landed at the worst possible point in Anthropic’s IPO timeline. The company confidentially filed its S-1 with the SEC on June 1, targeting an October 2026 Nasdaq listing with Goldman Sachs, JPMorgan, and Morgan Stanley as lead underwriters, in an offering expected to raise more than $60 billion. Anthropic closed a $65 billion Series H-1 round in late May at a $965 billion post-money valuation, and annual recurring revenue had reached roughly $47 billion in early June. A national security cloud over the company’s flagship models, even a brief one, is exactly the kind of disclosure risk that underwriters and institutional investors scrutinize in an S-1 process.

The G7 lunch in Évian-les-Bains on June 17 appears to have been the turning point. Amodei and DeepMind’s Demis Hassabis jointly proposed a coalition framework that would have democracies coordinate frontier AI trade and standards under US leadership, with China excluded. Trump described Amodei as smart and personable afterward, a notable shift from the February Truth Social framing that had characterized Anthropic’s leadership as ideologically hostile to the Defense Department. That the relationship was repaired through a geopolitical pitch rather than a technical rebuttal tells you something about how these designations get made and unmade.

The Defense Production Act remains available to Trump, who told Axios he could invoke it but sees no current need. Whether the Pentagon’s supply-chain risk designation is formally rescinded before Anthropic’s expected October listing will likely determine how much of this episode survives into the S-1’s risk factors.