Can US Chip Controls Keep China From Building Frontier AI?
November 15, 2026
Chinese researchers and technology companies are increasingly using distillation—training models on answers or demonstrations from more capable systems—to improve domestic AI, potentially easing the impact of US chip restrictions.
The debate intensified after Anthropic accused DeepSeek, Moonshot and MiniMax of using 24,000 fraudulent accounts to extract capabilities from Claude; the article says distilled models may lack safeguards and could create risks for critical infrastructure.
A Reuters review of more than 80 Chinese academic papers and patents found researchers associated with military and security institutions using American AI models for work including surveillance, software analysis and military applications.
At July meetings in Chengdu, US officials raised concerns with Chinese regulators, who called distillation standard practice and said they were open to addressing intellectual property theft but made no specific commitments.
Some Chinese developers have asked the Ministry of Industry and Information Technology to clarify when routine use of foreign models crosses into organized data collection, and the issue could arise around the November 18–19 APEC summit in Shenzhen, where Donald Trump and Xi Jinping are expected to meet.
A Chainalysis survey of 100 AI proxy providers that resell access to overseas models found that 40% accept cryptocurrency, mostly stablecoins, alongside payment options such as WeChat and AliPay.
These services route queries through intermediaries to conceal their origin and pool model access, while vendors use methods such as account farming and model swapping to obtain access.
Chainalysis says blockchain analysis can help investigators trace links among users, proxy operators, payment processors and exchanges, though the market can replace disrupted vendors and is difficult to dismantle.
Chainalysis observed that payments from wallets it suspects are China-linked rose 55% above baseline during the first three days after US frontier-model releases, then fell 21% below baseline around day 13, a period it associated with subsequent Chinese open-weight model releases.
The report cautions that aggregated on-chain activity cannot establish what particular payments funded or prove individual users’ motives.
Ben Buchanan argues that computing power—not money, talent or electricity—is China’s main constraint in developing frontier AI, and that US export controls have kept Chinese labs from training a frontier model on domestically made chips.
Buchanan says export controls have made China’s path to advanced chips harder, including through restrictions on Dutch extreme-ultraviolet lithography equipment, but acknowledges that US enforcement has failed to stop chip and chip-die smuggling.
Summary based on 30 sources
Get a daily email with more Tech stories
Sources

Yahoo News • Nov 17, 2026
Exclusive-China taps US AI models to train homegrown rivals as chip curbs bite
ET BrandEquity • Nov 17, 2026
China taps US AI models to train homegrown rivals as chip curbs bite
Economic Times • Nov 17, 2026
China takes a page from US AI playbook to build cheaper domestic rivals as chip curbs bite: Report
Deutsche Welle • Nov 16, 2026
Nvidia scandal: China's chips dilemma, US rivalry shape AI race