The U.S. State Department has issued a global alert following claims that Chinese companies, including AI startup DeepSeek, are allegedly involved in stealing intellectual property from U.S. AI labs. According to a diplomatic cable seen by Reuters, the alert aims to warn other nations about the security risks associated with using AI models distilled from U.S. proprietary models.
Distillation is a method used to train smaller AI models using outputs from larger, more sophisticated ones, often to reduce training costs for new AI systems. DeepSeek, a rising Chinese player in AI, recently drew attention with its low-cost AI model. The company granted preview access to its latest model, customized for Huawei chip technology, underscoring China’s increasing independence in AI development.
Despite requests for comments, neither the State Department nor the Chinese Embassy in Washington responded immediately. The cable also mentioned other Chinese AI firms, Moonshot AI and MiniMax, which also declined to comment.
The situation has intensified the technological competition between the U.S. and China, with the White House echoing similar concerns. In response, the Chinese Embassy dismissed the U.S. claims as unfounded, emphasizing its dedication to protecting intellectual property rights.
The cable, dated and distributed last Friday, encourages U.S. diplomatic staff worldwide to engage their foreign counterparts about worries regarding adversarial extraction of U.S. AI models. It also states that a separate message has been directed to Beijing, proposing further direct discussions.
Highlighting the urgency of the matter, the cable details the potential consequences of such unauthorized activities. AI models generated from unauthorized distillation efforts might perform similarly to the originals on some benchmarks but often lack comprehensive performance and essential security protocols. Moreover, these models may omit mechanisms that ensure neutrality and truthfulness.
Earlier this year, OpenAI had alerted lawmakers about DeepSeek’s alleged attempts to replicate its ChatGPT model and those from other top U.S. AI firms, potentially for its own development purposes. This growing concern over intellectual property theft comes at a politically sensitive time, just before a planned meeting between U.S. President Donald Trump and Chinese President Xi Jinping in Beijing, possibly reigniting tensions after a recent easing of hostilities.
Readers can access the original article here: [https://www.cnbc.com/2026/04/25/us-global-warning-alleged-china-ai-theft.html](https://www.cnbc.com/2026/04/25/us-global-warning-alleged-china-ai-theft.html)
