US aims to curb China's AI, but American models still reach its firms
Translated from Polish, summarized and contextualized by DistantNews.
At a glance
- US tech giants OpenAI and Google are reportedly supplying advanced AI models to Chinese technology firms blacklisted by the Pentagon, despite US efforts to curb China's AI development.
- These firms, including subsidiaries of Alibaba, Baidu, and Tencent, are allegedly collaborating with the Chinese military, creating a loophole in US export controls.
- While OpenAI suspended access for some Alibaba-linked users due to concerns about misuse, the broader US strategy struggles to prevent Chinese companies from accessing cutting-edge AI software.
US technology leaders OpenAI and Google are continuing to provide advanced artificial intelligence models to Chinese tech giants that have been placed on a Pentagon blacklist, according to a Financial Times report. This situation highlights a significant gap in Washington's strategy to slow China's progress in AI development.
The American companies have confirmed to the Financial Times that they are supplying AI services to Singapore-based subsidiaries of Alibaba, Baidu, and Tencent. The US government accuses these Chinese conglomerates of ties to the People's Liberation Army, raising concerns about the dual-use nature of AI technology.
Following inquiries from the Financial Times, OpenAI stated it suspended API access for users linked to Alibaba last month, citing concerns over the illicit use of its AI tools. Although such sales are currently legal, the revelations have intensified calls for stricter US regulations on AI models, mirroring existing restrictions on the export of chips used for training powerful AI systems.
The US government has taken steps to control access to specific advanced AI models, such as Anthropic's Mythos and Fable, and OpenAI's GPT-5.6. However, it has not implemented a comprehensive ban on China-based entities, including those on a congressional blacklist for alleged links to the Chinese military, from accessing state-of-the-art AI software. Chris McGuire, a technology and security expert at the Council on Foreign Relations, criticized the lack of export control measures, stating, "The Trump administration constantly says we have to beat China in AI, but the problem is they have done nothing on export controls, which is the real tool we have to slow China down."
OpenAI's recent action against Alibaba users stemmed from suspicions of "model distillation," a process where AI-generated outputs are used to improve competing systems. OpenAI maintains it does not permit access to its models within China but acknowledged allowing "certain companies" with Chinese ownership or operations to use its tools in regions where it can enforce safeguards and monitor for distillation. Google stated its AI services are available in Hong Kong and Singapore, subject to terms of use, including a ban on distillation. However, the company conceded that geographical sales restrictions alone are insufficient to mitigate distillation risks, as sophisticated adversaries can easily circumvent them. Alibaba has challenged its inclusion on the Pentagon's blacklist in a US court, calling the decision "arbitrary and capricious."
The Trump administration constantly says we have to beat China in AI, but the problem is they have done nothing on export controls, which is the real tool we have to slow China down.
Originally published by Rzeczpospolita in Polish. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.