US intelligence agencies accuse six Chinese AI companies of systematic technology theft
United States intelligence and law enforcement agencies on Tuesday formally accused six Chinese artificial intelligence companies of systematically copying technology from American AI firms, issuing a joint advisory that could strain diplomatic relations ahead of Chinese President Xi Jinping's scheduled visit to Washington on September 24.
The National Security Agency, Cybersecurity and Infrastructure Security Agency and Federal Bureau of Investigation named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI in the advisory, alleging the companies extracted intellectual property from US AI laboratories through a technique known as distillation. According to the agencies, these activities have been ongoing since at least late 2024, with Chinese companies extracting billions of tokens from American frontier AI models.
What is distillation?
Distillation is a widely-used machine learning technique where a smaller student model learns from a larger teacher model by training on the teacher's outputs, reducing computational costs and enabling deployment on resource-constrained devices. While the technique is considered legitimate when companies apply it to their own models, it becomes controversial when used at scale to extract capabilities from third-party proprietary models without authorization, according to industry experts.
Specific allegations against Chinese firms
The advisory alleged that DeepSeek distilled from Claude, ChatGPT, Gemini and Grok models to train its R1 and R3 systems, while Moonshot AI used millions of queries to train its Kimi-K2 and Kimi K3 models. The Chinese companies targeted AI models from Anthropic, OpenAI, Alphabet's Google and SpaceX, US officials said.
Diplomatic implications
The accusations come as President Trump prepares to host Xi Jinping later this month, following Trump's state visit to China in May where the Chinese leader was invited to the White House. The US and China are also preparing to hold their first official bilateral AI safety dialogue since Trump's second term, tentatively scheduled for mid-September and expected to be led by US Treasury Secretary Scott Bessent, though a White House official denied a meeting is currently planned.
US officials said the Chinese companies engaged in the distillation activities "likely with Chinese government awareness," adding that the practice not only reduces research and development costs for Chinese AI firms but also enhances China's "military and cyberattack capabilities that could be used against the US and our allies."
Escalating tensions
The allegations represent an escalation of accusations that have been building throughout 2026. In February, OpenAI warned US lawmakers in a memo to the House Select Committee on China that DeepSeek was using sophisticated distillation methods and obfuscated techniques to evade OpenAI's defenses. On April 23, a White House memorandum signed by Michael Kratsios accused foreign entities, primarily Chinese, of engaging in industrial-scale distillation campaigns of US-developed AI models.
In July, Reuters reported that Chinese military researchers have used outputs from leading US AI models developed by OpenAI and Anthropic to train domestic AI systems for defense applications, based on a review of more than 80 Chinese academic papers and patents.
The Chinese embassy in Washington did not immediately respond to a request for comment about the distillation allegations.
Officials characterized China's actions as "aggressive, malicious and targeted distillation activities at an industrial scale," marking one of the strongest public condemnations of Chinese AI development practices by the US government.
