Enterprise

Anthropic meets White House over Fable 5 export ban; no resolution as cyber experts revolt

Senior Anthropic staff met Commerce Department officials Monday over an unprecedented order that forced the AI company to pull Fable 5 and Mythos 5 offline worldwide. Talks ended without resolution.

By Asher Glick · Enterprise software desk · 2026-06-16 · 3 min read

Senior Anthropic technical staff sat down with Commerce Department officials on Monday to contest a Friday export-control directive that forced the company to pull Fable 5 and Mythos 5 offline for hundreds of millions of users worldwide. Per a person familiar with the planning, the meeting ended without resolution.

The directive landed at 5:21 p.m. ET on Friday. It ordered Anthropic to suspend access to both models for any foreign national “whether inside or outside the United States,” a scope so broad it captured Anthropic’s own foreign-national employees and effectively required disabling both models for every customer. Access to the company’s other models is unaffected.

What actually triggered the order depends on who’s telling the story, and the two official accounts don’t reconcile.

Fortune reports that Amazon CEO Andy Jassy raised concerns with senior administration officials on Thursday, after Amazon researchers used prompts to elicit restricted cyberattack information from a Mythos-class model. Amazon has $8 billion already invested in Anthropic and committed up to $25 billion more in April, tied to commercial milestones, a financial entanglement large enough that any Amazon-sourced security alarm doubles as commercial signaling.

White House AI adviser David Sacks told a different version on X: a trusted partner of both Anthropic and the government identified a jailbreak in Fable 5, the administration asked CEO Dario Amodei to fix or withdraw the model, and Amodei refused.

Anthropic’s read is narrower. Executives say they reviewed a demonstration of the technique and found it surfaced only a small number of previously known, minor vulnerabilities, flaws other publicly available models can discover without any jailbreak. Recalling a commercial model on those grounds, they argue, would halt all frontier deployments if applied industry-wide.

The cybersecurity community has lined up behind that framing. Dozens of industry veterans signed an open letter arguing the order “has taken the best models away from defenders.” Researcher Katie Moussouris, who reviewed a private copy of the Amazon paper at Anthropic’s request, said the behavior “cannot meaningfully be fixed, and any attempt would only weaken the model for defense.”

The financial backdrop sharpens everything. Anthropic confidentially filed for a public listing earlier this month at a $965 billion valuation. The Pentagon designated the company a “supply chain risk” in early March, a label Anthropic is challenging in federal court. The export order arrives on top of a designation fight, in the middle of an IPO process, with the company’s largest backer credited as the proximate cause. That sequence isn’t security policy operating in isolation. It’s security policy operating inside a commercial system that has every incentive to use it.

Sources