U.S. Agencies Accuse Six Chinese Firms of AI Copying

The FBI, NSA and CISA accused six Chinese AI companies—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI—of conducting unauthorized, industrial-scale distillation of U.S. frontier models since at least late 2024. The agencies said the firms extracted billions of tokens through millions of requests involving models from Anthropic, OpenAI, Google and xAI, allegedly using fraudulent or shared accounts, bulk subscriptions, cloud services, proxy “transfer stations” and prompt-injection techniques to evade geographic restrictions, usage limits and detection. Anthropic separately alleged that some companies routed millions of real user queries, potentially including sensitive information such as locations and passwords, through intermediaries to support model distillation. U.S. officials said the activity could reduce development costs and timelines, aid systems including DeepSeek’s R1 and V3 and Moonshot’s Kimi models, and likely occurred with Chinese government awareness, while noting that authorized distillation is legitimate. China rejected the accusations as groundless and defended its AI progress as domestically developed, as tensions rise ahead of expected U.S.-China discussions on AI governance and broader technology competition; U.S. officials said sanctions or Entity List designations could follow.