Overview
- The Commerce Department issued an emergency export‑control directive that barred any foreign national from using Anthropic’s Fable 5 and Mythos 5, and Anthropic disabled both models worldwide to comply.
- The order was delivered in mid‑June after private tests reported that targeted prompts could bypass model safeguards and surface restricted cybersecurity information.
- Amazon researchers and other private parties reportedly flagged the issue to U.S. officials, which helped trigger high‑level government calls with Anthropic about the vulnerability.
- Anthropic says officials provided only verbal evidence of a narrow, non‑universal jailbreak and disputes the severity of the finding while sending senior staff to Washington to negotiate a fix and restoration.
- The move is an unprecedented use of export controls on cloud AI, disrupting global customers and foreign employees, complicating Anthropic’s IPO plans, and raising pressure for sovereign or multi‑source AI strategies.