Overview
- Anthropic disabled its Fable 5 and Mythos 5 models for all users after the Commerce Department issued an export‑control directive on June 12–13 that barred access by any foreign national.
- U.S. officials cited a reported jailbreak that lets the models read code and produce exploit‑grade fixes, a flaw that Amazon researchers and others said could be used to create cyberattacks.
- Anthropic sent technical and senior staff to Washington to negotiate with White House and Commerce officials and President Trump said talks with the company were "going fine" as European Commission officials separately met Anthropic in San Francisco to seek tightly controlled EU access.
- Experts say applying export controls to cloud inference is unprecedented and legally unclear because firms cannot reliably block users by nationality in real time and the rulebook does not map cleanly onto remote AI services.
- The move puts near‑term commercial and IPO prospects for Anthropic at risk, creates market uncertainty for U.S. AI firms, and could push allies and customers toward alternative suppliers or open‑source models that lack the same safety controls.