US State Department Warns APAC Allies on Chinese AI Distillation

On April 24, 2026, the US State Department sent a cable to every diplomatic and consular post it operates around the world. The message was pointed: warn your host governments that Chinese AI companies are stealing American technology.

The cable singled out three Chinese AI firms by name: DeepSeek, Moonshot AI, and MiniMax. It described a practice called "distillation," where a smaller AI model is trained using outputs from a larger, more expensive one. The US government says Chinese companies have been running these campaigns secretly, against American AI labs, without permission.

For Asian tech leaders, this is no longer a Washington-Beijing spat to observe from a distance. The US is now formally asking allied governments to take sides.

What the Cable Actually Says

The State Department cable was direct about its goals. It aimed to "warn of the risks of utilising AI models distilled from US proprietary AI models, and lay the groundwork for potential follow-up and outreach by the US government."

It went further than a simple IP complaint. The cable claimed that distillation campaigns "deliberately strip security protocols from the resulting models and undo mechanisms that ensure those AI models are ideologically neutral and truth-seeking." In other words, Washington is framing this not just as theft, but as a potential security risk embedded in the technology itself.

The White House also made similar accusations publicly this week, though the cable itself had not previously been reported. The State Department did not immediately respond to a request for comment.

Beijing Pushes Back, and Keeps Building

China's response was firm and immediate. The Chinese Embassy in Washington called the allegations "groundless" and characterized them as "deliberate attacks on China's development and progress in the AI industry."

DeepSeek did not comment on the cable. The company has previously said that its V3 model relied on publicly available web data and did not intentionally use data generated by OpenAI.

On the same day the cable surfaced, DeepSeek launched a preview of its new V4 model, built specifically to run on Huawei chips. The timing was either coincidence or a pointed signal: China's AI ambitions are not slowing down, and the country is investing in hardware independence to match.

That trajectory is backed by real numbers. Chinese AI models went from handling roughly 1% of global AI workloads in late 2024 to an estimated 30% by end of 2025. The Alibaba Qwen model family alone has accumulated over 700 million downloads worldwide.

The Awkward Position APAC Governments Are In

Asian governments are not speaking with one voice on this. Taiwan has banned DeepSeek across all government agencies, critical infrastructure, public schools, and government-funded organizations. The concern there is data privacy and cross-border transmission risk.

Hong Kong has gone in a different direction. Its government built HKGAI V1, its first local AI model, directly on DeepSeek's open-source parameters, supporting Cantonese, Traditional Chinese, English, and Putonghua.

Singapore's minister for digital development welcomed DeepSeek publicly as a cost-reducing innovation. Around the same time, Singapore authorities arrested three individuals for facilitating AI chip smuggling to China. Both things are true at once.

The Reputational Risk Is Already Priced In

For communications and business leaders in Asia, the risk here is not just regulatory. It is reputational.

AI vendor associations have become the fastest-growing risk disclosure among large public companies. In 2023, just 12% of S&P 500 companies flagged AI reputational risk in their filings. By 2025, that figure had risen to 38%. Investors and boards are paying attention.

Any APAC organization that publicly associates itself with Chinese AI tools now faces scrutiny from two directions: US-aligned regulators and partners asking about distillation risks, and Chinese counterparts sensitive to any signal of alignment with Washington.

The legal picture adds one more layer of complexity. Experts at law firm Winston and Strawn note that AI model distillation, while potentially violating a company's terms of service, does not constitute clear IP theft under existing law. There is no international legal framework to enforce these claims across borders. The US government's diplomatic framing runs ahead of what courts can actually do.

The Stakes Going Into the Trump-Xi Summit

All of this is happening fewer than three weeks before President Trump is scheduled to visit President Xi in Beijing. The State Department cable and the White House's public accusations could raise tensions that a detente brokered last October had partially lowered.

Whether AI IP becomes a formal bargaining chip in those talks remains to be seen. But the signal sent to Asian governments is already clear: Washington expects its allies to take these concerns seriously.

For tech leaders in the region, the question is less about which side is legally right. It is about which position their organization can credibly defend, to regulators, partners, clients, and boards, as the pressure intensifies on both sides.