The United States has dispatched a diplomatic cable to its embassies and consulates worldwide, ordering diplomats to press host governments on the alleged theft of American artificial intelligence models by Chinese firms. The cable, sent on Friday, specifically instructs missions to raise concerns about the “distillation” of proprietary US systems, naming the start-ups DeepSeek, Moonshot AI and MiniMax. A separate note of protest has been delivered to Beijing. This marks a distinct escalation: Washington is no longer content with bilateral accusations but is now actively building a global coalition against what it frames as industrial-scale intellectual property theft.

Viewed from Washington, the logic is straightforward. White House technology adviser Michael Kratsios has asserted that the US possesses evidence of foreign entities, primarily in China, secretly using American models at scale. In Senate testimony, Helen Toner of Georgetown University’s Centre for Security and Emerging Technology described distillation as a common technique whereby a smaller “student” model is trained on the outputs of a more advanced “teacher” model, allowing developers to replicate sophisticated capabilities at a fraction of the cost. US officials argue this allows Chinese companies to free-ride on billions of dollars of American research investment, and the cable lays the groundwork for further restrictive measures.

Beijing has rejected the accusations as a “completely unfounded” smear campaign against China’s AI achievements. Foreign ministry spokesman Guo Jiakun urged Washington to “respect the facts,” and within hours of the White House allegations, DeepSeek announced the release of a new model, a timing that analysts in the region read as deliberately defiant. Yet behind the official rebuttal, there is unease. Analysts warn that a sustained US crackdown on distillation could trigger a sharp shake-out of China’s crowded AI sector, where many weaker start-ups depend heavily on copying teacher models. Within a year, they say, the less innovative firms could be forced out, leaving only a handful of well-resourced players.

Parallel to the diplomatic offensive, a second front is opening on investment flows. Chinese regulators, including the powerful National Development and Reform Commission, have quietly instructed a select group of AI start-ups to reject American investments in new funding rounds without explicit government approval. Moonshot AI and Stepfun are among those that have received such directives, while ByteDance is being pressed not to sanction secondary share sales to US investors. Viewed from London, this double movement—Washington warning the world about Chinese distillation while Beijing erects new financial barriers—represents a further, deliberate decoupling of the two AI ecosystems.

Each side is now reinforcing the other’s narrative. Washington points to distillation as proof that Chinese firms cannot be trusted with access to American technology. Beijing frames its investment curbs as a necessary defence against extraterritorial overreach. Analysts in London note that the distillation dispute is both technically genuine and strategically convenient. If the US succeeds in stigmatising the practice, it may force Chinese firms to develop wholly indigenous models more quickly, but it risks fracturing global AI governance standards. The next critical test will be the reaction of governments in the Global South, where cost-effective Chinese AI tools, often built on distilled models, are already in widespread use. Their readiness to heed Washington’s warnings may determine whether this diplomatic campaign succeeds or quietly stalls.