Chinese AI companies are systematically using output from leading American models to train their own competing systems, a practice known as distillation that has moved from academic experiments into mainstream industrial use, according to a Reuters review published November 17 of more than 300 Chinese academic papers, patents, corporate disclosures, recruitment notices and government documents.
The findings come as the United States tightens chip export controls intended to limit China's ability to develop advanced AI. On Monday, the Commerce Department restricted shipments of advanced chips to countries including Malaysia, Singapore, Thailand, Indonesia, and Brazil, and imposed new licensing requirements on overseas data centers supplying significant computing power to China-headquartered firms [6]. The rule was published days after Nvidia CEO Jensen Huang resigned amid reports of an affair with a suspected Chinese spy at Singapore-based cloud partner Megaspeed.
How Distillation Works #
Distillation uses a powerful "teacher" model to generate answers or demonstrate task completion, which developers then use to train a smaller "student" system. For Chinese developers facing computing-power constraints, "the attraction is in making one expensive query do repeated work," a former Alibaba engineer told Reuters [1]. An American model generates the answer to a difficult question once, allowing a Chinese team to retain useful examples and train a local system around them.
The process cannot reproduce a leading model's full capabilities or eliminate the need for advanced computing power, but it can help preserve selected skills in smaller systems that cost less to operate, potentially reducing reliance on American technology [1].
Documented Cases from Chinese Research #
Reuters identified several specific instances in Chinese publications:
-
November 2024 : Shanghai Jiao Tong University researchers selected about 5,000 answers from OpenAI's o1 model, translated them using GPT-4o mini, and assembled 10,750 bilingual training examples to further train Alibaba's Qwen2.5-72B model. The researchers cautioned that excessive reliance on distillation could stifle original innovation[1] .
-
April 2025 : Alibaba researchers described using OpenAI's GPT-4 and GPT-4o to generate training material in languages other than Chinese, then refining and scoring responses before training smaller Qwen models[1] .
-
February 2026 : Researchers from Shanghai Jiao Tong University and food-delivery company Meituan used GPT-4o to generate plans for coordinating AI agents working together on programming problems, helping train a smaller model to organize such tasks[1] .
Alibaba, Meituan, and Shanghai Jiao Tong University did not respond to Reuters requests for comment.
Geopolitical Friction #
US authorities publicly accused Chinese companies in September of "aggressive, targeted distillation of American AI models on an industrial scale." In July, Commerce Department officials raised concerns with Chinese regulators in Chengdu, warning that some activities could amount to intellectual-property theft, according to three people familiar with the discussions. Chinese officials defended distillation as a standard AI-development technique and said they were open to acting where conduct constituted IP theft, but made no concrete commitments, a US official said [1].
A US official familiar with the Chengdu discussions said Washington believes industrial-scale distillation is the main reason Chinese AI models have narrowed the gap with American rivals. Two people familiar with Chinese developers' operations disputed that assessment, saying distillation was only one factor alongside advances in engineering, algorithms, data and model design [1].
OpenAI's terms restrict using its output to develop competing models, subject to exceptions, while Anthropic similarly restricts training competing systems without permission [1].
Chinese Policy Encourages Distillation #
Chinese government documents do not instruct companies to distil American models but encourage distillation as part of Beijing's push to spread AI across the economy. A national AI+Manufacturing policy package published by eight Chinese government agencies in January includes guidance supporting model training, fine-tuning and distillation for industrial applications. Shanghai and Shenzhen have also encouraged techniques to reduce computing requirements [1].
The policies reflect Beijing's ambition to deploy smaller, cheaper AI systems across industries. Concerned by distillation's growing geopolitical sensitivities, Chinese AI developers have in recent months asked Beijing's industry regulator for guidance on the boundaries between ordinary use of foreign models and organized collection of output for training, according to one person familiar with the discussions [1]. China's Ministry of Industry and Information Technology did not respond to a request for comment.
US Response: Closing Compute Loopholes #
The distillation findings coincide with a broader US effort to close off-chip access routes. The Commerce Department's new rule requires overseas data centers providing significant compute volume to China-headquartered companies to obtain US export licenses within 90 days or face future restrictions. It also restricts chip shipments to data centers above a certain compute threshold [6].
Separately, the United States and nine allies—including Australia, India, Japan, South Korea, and the United Kingdom—signed the San Francisco Declaration on November 16 to coordinate protection of AI technologies and critical infrastructure under the Pax Silica initiative [2]. Germany, Sweden, the Netherlands and the European Commission are set to sign next week.
At the APEC summit in Shenzhen, where President Trump and President Xi are expected to meet November 18-19 [1], APEC chair Chen Xu said members had "reached an important consensus around AI-related issues" including "the safe deployment and responsible application of AI" [4]. The US chip restrictions announced Monday were framed as preventing Chinese companies from accessing advanced US AI chips through overseas data centers—a loophole Washington says could allow firms to circumvent existing export controls [4].
What Remains Unknown #
Reuters' review covers documents through early 2026; the current scale of distillation in production systems is not quantified. The effect of the new chip restrictions on distillation practices has not been assessed.
Sources #
1. [China taps US AI models to train homegrown rivals as chip ...](https://www.reuters.com/world/china/china-taps-us-ai-models-train-homegrown-rivals-chip-curbs-bite-2026-11-17) (reuters.com)
2. [US, allies team up to protect AI tech and infrastructure amid ...](https://www.reuters.com/world/china/us-allies-team-up-protect-ai-tech-infrastructure-amid-china-rivalry-2026-11-16) (reuters.com)
3. [Inside China's drive to build an AI-native generation](https://www.reuters.com/world/china/inside-chinas-drive-build-an-ai-native-generation-2026-11-16) (reuters.com)
4. [APEC focuses on safe, responsible AI use, China's ex-UN ...](https://www.reuters.com/technology/apec-focuses-safe-responsible-ai-use-chinas-ex-un-envoy-says-2026-11-17) (reuters.com)
5. [Trump says he might meet with North Korea's Kim Jong Un ...](https://www.reuters.com/world/trump-says-he-might-meet-with-north-koreas-kim-jong-un-china-2026-11-16) (reuters.com)
6. [US to curb China's access to AI chips as Trump heads to Asia](https://malaysia.news.yahoo.com/us-curb-chinas-access-ai-220054158.html) (malaysia.news.yahoo.com)