The NSA, CISA, and FBI have issued a joint cybersecurity advisory accusing Chinese AI companies of conducting "distillation activities at an industrial scale" to extract proprietary capabilities from American frontier models. The agencies specifically name Deep Seek, Moonshot AI, Alibaba, Mini Max, Step Fun, and Z.AI as extracting "billions of tokens across millions of exchanges/requests" from models including Anthropic's Claude, OpenAI's GPT, Google's Gemini, and xAI's Grok since 2024.

Distillation involves using outputs from powerful models to train newer ones. The advisory states Deep Seek used data from Claude, Gemini, GPT, and Grok models to train its R1 reasoning model released in early 2025. Moonshot AI allegedly extracted significant data from Anthropic's Fable to train Kimi K3 and used GPT-4o data for Kimi K2. Anthropic created Fable to bring capabilities from its Mythos cybersecurity model to the public.

The warning details mitigations US firms can implement against malicious distillation campaigns. American companies have long alleged Chinese firms copy their technologies. OpenAI previously banned accounts suspected of distilling its technologies, including Deep Seek. Anthropic also accused Deep Seek, Moonshot, and Mini Max of wide-scale distillation. Notably, OpenAI noted Elon Musk admitted to using OpenAI's output to train xAI models during cross-examination in his lawsuit against the company.