OpenAI and Google are supplying AI to Chinese groups on the Pentagon’s blacklist
OpenAI and Google have confirmed that they provide artificial intelligence services to Singapore-based subsidiaries of Alibaba, Baidu and Tencent. This legal practice has reignited the debate in the United States over export controls on AI technologies
OpenAI and Google have confirmed that they have provided artificial intelligence services to Singapore-based subsidiaries of Alibaba, Baidu and Tencent, three Chinese tech giants included on the Pentagon’s so-called ‘1260H’ list, which comprises companies accused by Washington of having links to the People’s Liberation Army. Although these transactions were legal, the affair has reignited the debate in the United States over whether to extend AI models restrictions similar to those already in place for the export of advanced semiconductors.
According to what the two companies confirmed to the Financial Times, OpenAI and Google provide access to their models via subsidiaries located in countries such as Singapore and Hong Kong, where these services are available and subject to their respective terms of use.
OpenAI has explained that it does not allow direct access to its models from within China, but does authorise “certain companies” based in or owned by Chinese entities to use its tools in countries where it is possible to implement security measures and monitor any misuse. The company also announced that last month it had suspended access to its APIs for users affiliated with Alibaba after detecting suspected cases of ‘distillation’ – the technique whereby developers exploit the outputs of one model to train a competing one. The matter has also been reported to the US government.
Google has stated that its AI services are available in Hong Kong and Singapore in accordance with its terms of service, which expressly prohibit the practice of distillation. The company has, however, emphasised that geographical restrictions alone are not sufficient to prevent this type of activity, as they can be circumvented by particularly sophisticated actors.
US legislation currently restricts access to certain cutting-edge models, including OpenAI’s GPT-5.6 and Anthropic’s Mythos and Fable models, but does not impose a blanket ban on the supply of AI software to Chinese companies, even when they appear on the Pentagon’s blacklist. Several experts view this as a loophole in the US strategy to curb the development of Chinese artificial intelligence.
