Overview
- A U.S. government memorandum found deliberate, industrial-scale campaigns to 'distill' capabilities from U.S. frontier models, prompting Washington to limit foreign access to some systems.
- The U.S. ordered Anthropic to suspend foreign-national access to advanced models, forcing a temporary global disable of those models and later allowing limited vetted access for some defenders.
- Anthropic added a client-side marker to its Claude Code to detect suspected distillation and has since removed the marker while saying it will strengthen safeguards.
- Alibaba has placed Claude Code on a high-risk list and told employees to stop using Anthropic tools after Anthropic accused several Chinese labs of large-scale distillation attempts.
- Chinese regulators have held talks with major firms about possible export-style limits and using national-security law against AI theft, a debate that could slow the spread of open-weight models and shift global AI adoption.