Back
Anthropic works with White House on AI security rules after model takedown order

Anthropic works with White House on AI security rules after model takedown order

After an emergency 90-minute order to remove the Fable 5 and Mythos 5 models, Anthropic is in talks with White House and Trump administration officials to address security concerns and broader oversight issues.

Fact Check
Every core element of the claim is corroborated by multiple authoritative independent sources. Fortune and Politico confirm Anthropic received roughly 90 minutes' notice/ultimatum to pull Fable 5 and Mythos 5, and Wired confirms the models were disabled to comply with a US government export-control order. Cryptobriefing (citing NYT) and odaily report that Anthropic is in active talks with White House/Trump administration officials, including named executives Tom Brown and Sarah Heck, to address security concerns and broader oversight—matching the claim. The only imprecision is the claim's phrasing 'model takedown order': technically it was a Commerce Department export-control directive barring foreign-national access that Anthropic then enforced as a full global disable. This does not undermine the substance.
Summary

Anthropic is working more closely with the White House on AI security rules after an emergency takedown order covering the Fable 5 and Mythos 5 models heightened internal concern about governance and IPO-related risks. The New York Times reported that management was given 90 minutes to remove the models, while some employees were unsettled by shifting explanations that alternated between "national security" and "security vulnerabilities." People familiar with the matter said Anthropic told U.S. Commerce Secretary Lutnick it would cooperate more closely with the White House, with co-founder and chief compute officer Tom Brown and public policy head Sarah Heck leading talks with Trump administration officials as of Thursday. No firm timeline has been set for a full resolution.

Terms & Concepts
  • IPO: initial public offering of shares
  • AI security rules: Policies and safeguards that govern how artificial intelligence systems are developed, deployed and monitored.
  • security vulnerabilities: weaknesses that can be exploited