Senior technical staff from Anthropic PBC left the Commerce Department on Monday without an agreement, leaving Friday’s export-control directive against Claude Fable 5 and Claude Mythos 5 in force and the company’s two flagship models offline worldwide.

The directive landed at 5:21 p.m. EDT on June 12, citing “national security authorities” and barring foreign nationals, including Anthropic’s own non-U.S. employees, from accessing either model. Rather than partition access by nationality, Anthropic disabled the models globally and began contesting the order through back channels.

The proximate trigger is a jailbreak report Amazon submitted Thursday evening, according to Axios. Anthropic’s blog post calls the finding “narrow” and reproducible against other publicly available systems, including OpenAI’s GPT-5.5. Katie Moussouris, CEO of Luta Security, who reviewed the report at Anthropic’s request, told Axios the government response “seems way out of line with what’s actually in the research report.”

The commercial subtext is hard to miss. Amazon CEO Andy Jassy was among the executives who raised concerns with senior administration officials, a White House official confirmed to CNBC. Amazon has invested $8 billion in Anthropic and committed up to $25 billion more, making it the lead patron of a company whose models its own report helped take offline.

This isn’t the first front. In February the Department of Defense designated Anthropic a supply-chain risk, a finding the company is challenging in federal court. Friday’s directive layers an export-control fight on top of an unresolved procurement one.

Allied capitals are already pushing back. At the G-7 summit in Évian-les-Bains, European diplomats raised “trusted partner” carve-outs with Commerce Secretary Howard Lutnick, the Financial Times reported. The 2019 Huawei entity-listing pattern is repeating itself in reverse: this time the export control points inward, and the company on the wrong side of it’s American.

Sources