U.S. claims China built top AI systems by cloning American models

The FBI, NSA and CISA issued their most detailed accusation yet, claiming that six major Chinese AI companies have systematically copied U.S. AI models through large‑scale distillation campaigns since 2024. According to the alert, Chinese developers mined models such as ChatGPT, Claude, Gemini and Grok to build their own systems, allegedly with Chinese government awareness. Distillation is a common technique in AI, but using it to replicate proprietary systems violates the terms of service of American companies. The report lists specific capabilities Chinese models supposedly extracted, including legal reasoning, agentic functions and coaching‑style assistance. China dismissed the allegations as groundless, insisting its AI progress comes from domestic innovation and calling for global cooperation instead of accusations. U.S. officials argue that covert industrial‑scale distillation allows Chinese firms to bypass safety guardrails and piggyback on American research. The alert urges U.S. companies to strengthen user verification, share intelligence on suspicious activity and degrade responses to malicious queries.

Read the full story on NBC News →