In an official joint statement, the US accuses Chinese AI firms of engaging in malicious, aggressive, and targeted copying of proprietary American artificial intelligence technologies. Federal law enforcement and intelligence agencies disclosed that six China-based enterprises have systematically leveraged model distillation—a methodology where smaller, secondary models are trained on outputs from larger, highly complex architectures—to bypass expensive research and development costs. Officials noted that these operations are occurring with the probable knowledge of the Chinese government and specifically target leading American labs including Anthropic, OpenAI, Alphabet’s Google, and SpaceX.
The joint advisory emphasizes that this industrial-scale distillation not only grants commercial shortcuts to domestic developers like DeepSeek, Moonshot AI, and Alibaba, but also actively enhances foreign military and cyberattack capabilities. US security officials warned that outputs extracted from frontier American models are being integrated into defense applications, directly threatening national security and allied defense frameworks.
Geopolitical Friction Ahead of Presidential Summit
These formal allegations arrive during a delicate diplomatic period for the Trump administration, which is preparing for Chinese President Xi Jinping’s upcoming visit to the United States in late September 2026. Bilateral working groups from both nations are also scheduling mid-September talks dedicated specifically to evaluating artificial intelligence safety risks. Similar intellectual property concerns were raised earlier in April prior to executive visits to Beijing, underscoring persistent tension surrounding digital sovereignty and frontier model security.
The Chinese embassy in Washington did not immediately respond to requests for comment regarding the joint advisory. The confrontation highlights expanding regulatory, economic, and defense friction over synthetic training data, frontier weights, and global technology market share.