Anthropic disclosed that three Chinese AI labs extracted 16 million prompts through 24,000 fake accounts, with MiniMax redirecting half its traffic to new Claude models within 24 hours. The timing—11 days after OpenAI's similar memo to Congress—highlights how API access undermines export controls on AI chips. The technical threat is real, but so is the coordinated policy messaging.

#AIPolicy #ExportControls #AIGovernance

https://www.implicator.ai/distillation-attacks-on-claude-are-real-so-is-the-lobbying-campaign/