Overview
- The Commerce Department told Anthropic on Friday to block non‑U.S. nationals from using Fable 5 and Mythos 5 after security researchers reported prompts that can make the models surface software‑vulnerability information.
- Anthropic responded by disabling both models for all users to comply with the export directive, saying the reported bypass is narrow and that evidence given by officials was verbal.
- Amazon researchers flagged the issue to U.S. officials and CEO Andy Jassy briefed the administration, which held high‑level calls with Anthropic before the export controls were issued.
- The move affects customers and partners worldwide, including large Indian IT firms and European agencies, and renewed calls in Brussels for greater AI sovereignty and local compute capacity.
- Experts say the order sets a new precedent by using export controls on live cloud AI services and could reshape how governments, companies, and researchers share and regulate frontier models.