US government accuses six Chinese AI firms of stealing US AI models
The US National Security Agency and FBI accused six Chinese AI companies on September 8 of systematically stealing US AI models through 'distillation'—a technique where new models learn from existing ones. Anthropic reported that its Claude AI was used without authorization and that Chinese military information was transferred without permission by the accused firms.
What happened
The US National Security Agency and Federal Bureau of Investigation issued a statement on September 8 accusing six Chinese AI firms of systematically misusing US AI models to train their own systems through a technique known as 'distillation.'
The agencies specifically named DeepSeek and Moonshot AI among the Chinese companies engaged in what they characterized as 'industry-scale malicious, unethical, and targeted distillation activity' against US AI companies.
Reported impact
Anthropic reported that Claude AI was used without authorization by these actors to train Chinese AI systems, with Chinese military information also being transferred without permission.
Index analysis
Distillation is a technique whereby new AI models leverage the knowledge of existing models to improve their performance. The US government views the Chinese firms' unauthorized use of US models for this purpose as theft of intellectual property and competitive advantage.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.