US accuses Chinese AI firms of copying leading American models

The US National Security Agency (NSA), Cybersecurity and Infrastructure Security Agency (CISA), and Federal Bureau of Investigation (FBI) have issued a joint official statement. As reported by Ars Technica, American authorities accuse six Chinese companies of stealing and copying the capabilities of leading US artificial intelligence models on a massive scale.

The joint document published by the security agencies specifically names six Chinese AI companies: DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI. According to security services, these organisations have been conducting industrial-scale attacks since at least late 2024 aimed at illegally extracting and appropriating functionality from American flagship models.

Targets and economic benefits

US agencies stated that targets included versions of leading American AI systems such as Claude, GPT, Gemini, and Grok. The report emphasises that the Chinese companies likely acted with the full knowledge and awareness of the Chinese government.

Security services explained that the process known as distillation — training a new model by extracting capabilities from an existing one — allows Chinese firms to dramatically reduce development and training timelines for their AI models.

In addition to saving significant time, this method enables Chinese companies to avoid billions of dollars in financial expenditure necessary to build and train next-generation frontier models from scratch.

This joint accusation from US intelligence and security agencies highlights the expanding scale of global competition and technological confrontation in artificial intelligence between Washington and Beijing.