At 5:21 p.m. ET on Friday, June 12, a letter from Commerce Secretary Howard Lutnick landed in Dario Amodei's inbox citing the Export Control Reform Act of 2018 and barring foreign-national access to Anthropic's two strongest models. By the end of the night, Anthropic had pulled Claude Fable 5 and Claude Mythos 5 globally. Fable 5 had been public for three days.

This is the first time the Commerce Department has reached for the ECRA's emerging-tech authority, according to an export-control expert cited by Reuters. The instrument was written in 2018 for a hypothetical future in which a U.S. frontier capability needed walling off from "countries of concern." Lutnick's letter, per the Globe and Mail's account, gestures at exactly that frame: military intelligence users in China, Russia, or elsewhere potentially deploying the models.

The supporting evidence is thinner than the authority is heavy. Anthropic says the government provided only verbal description of a "potential narrow, non-universal jailbreak," which the company characterizes as asking the model to read a specific codebase and fix any software flaws. That capability, Anthropic notes, is widely available from other models including OpenAI's GPT-5.5, and is used daily by defensive security teams. Mythos itself, restricted since April under Project Glasswing to roughly 50 vetted organizations including Amazon, Apple, Google, Microsoft, CrowdStrike and OpenAI, has been finding flaws in every major codebase its operators point it at.

Anthropic's response is unusually direct: "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers." Fable 5 went through thousands of hours of red-teaming with the U.S. government, the UK AISI, and third parties before launch. Vals AI benchmarks rated it the most capable publicly available model at release. Claude Opus 4.8 remains online.

The political backdrop is the part the safety framing doesn't quite cover. In February, Trump ordered federal agencies to stop using Anthropic after the company refused Pentagon contract language permitting use "for any lawful purpose." In March, the Pentagon labeled Anthropic a supply chain risk. Earlier this month, the company confidentially filed for an IPO at a $965 billion valuation.

Negotiations are now active. Reuters reports Anthropic's technical staff meeting Commerce officials almost daily; National Cyber Director Sean Cairncross joined a working-level session Monday. More than 80 cybersecurity executives, including signatories from Nvidia and Adobe, published an open letter backing Anthropic over the weekend.

What sits underneath is a question the 2018 statute was never stress-tested against: whether a verbally described jailbreak, in a capability available elsewhere, clears the bar for the heaviest export tool in the federal kit. The answer Commerce gives here becomes the template every frontier lab is now reading against its own roadmap.

U.S. forces Anthropic to disable Fable 5 and Mythos 5 worldwide over a verbal jailbreak claim

Sources