Overview
- The government told Anthropic to bar access to Fable 5 and Mythos 5 for foreign citizens, a restriction the company said arrived on Saturday and led it to disable those models for all customers to ensure compliance.
- Officials flagged a reported 'jailbreak' technique that could let the models find software security flaws and be misused in cyber operations, a concern that several outlets say followed reports from private firms and government cyber agencies.
- Anthropic said it reviewed a demonstration of the technique and found the flaws were limited, but argued that the government action did not follow the transparent, data‑driven process it supports and that other Claude models remain available.
- The move widens an existing clash with U.S. authorities after the Pentagon labeled Anthropic a supply‑chain risk and the company sued to challenge that designation, and it comes days after Anthropic publicly launched Fable 5 while filing confidentially with the SEC.
- The shutdown immediately disrupts commercial availability and customer access, raises questions about how firms can enforce nationality‑based export controls on cloud AI, and is likely to prompt legal and regulatory fights over oversight and model rollout.