The US Department of State has ordered a diplomatic campaign to draw attention to the efforts of Chinese AI companies to steal intellectual property from US laboratories.
According to a diplomatic telegram first reported by Reuters, the administration is increasingly concerned that companies, including the well-known startup DeepSeek, are using illegal "distillation" techniques to replicate the capabilities of advanced U.S. systems.
The telegram, sent to diplomatic and consular missions on Friday, aims to warn foreign partners about the risks associated with AI models that are based on U.S. proprietary research.
Distillation, the process of training smaller, cost-effective models by using the results of more powerful systems, is a legitimate industry practice when used in isolation.
However, the State Department and the White House claim that Chinese participants are conducting "industrial-scale" campaigns to circumvent security protocols and replicate U.S. technological breakthroughs at a fraction of the development cost.
Tensions rise on eve of Beijing summit
The aggressive posture marks a significant hardening of the Trump administration's approach to the ongoing AI arms race, coinciding with similar accusations made by the White House earlier this week.
The directive instructs U.S. diplomats to raise questions regarding the "extraction and distillation" of U.S. models with an official demarche request sent directly to Beijing.
The move comes just weeks before President Donald Trump is scheduled to meet with Chinese President Xi Jinping in Beijing, threatening to reignite tensions in a technology rivalry that experienced a brief thaw last October.
Chinese authorities have dismissed the allegations as "baseless" and "pure slander." The diplomatic cable suggests that the administration is preparing to move beyond rhetoric, laying the groundwork for potential follow-up measures and coordinated efforts with international partners.
Market and Competitive Risks
For the broader AI sector, these revelations highlight the growing "war premium" that is now being applied to intellectual property security.
Anthropic, OpenAI, and other leading U.S. labs have previously sounded the alarm, alleging that firms like DeepSeek, Moonshot AI, and MiniMax have been using rogue proxy accounts to siphon proprietary reasoning traces from top models.
As the competition for AI dominance intensifies, the U.S. government appears to be prioritizing the protection of "ideologically neutral and truth-seeking" security mechanisms, which it claims are being systematically circumvented by foreign distillation campaigns.