OpenAI and Google have confirmed that they supplied advanced artificial intelligence services to Singapore-based subsidiaries of Alibaba, Baidu, and Tencent, companies that the US government has accused of having links to China’s military.
The disclosures have renewed debate over whether Washington should tighten export controls on advanced AI models in addition to restrictions on semiconductor exports.
The companies told the Financial Times that the services were provided legally through overseas subsidiaries.
However, the arrangements have highlighted what critics describe as a gap in US efforts to limit China’s access to cutting-edge AI technologies.
OpenAI suspends Alibaba-linked API access
OpenAI said it suspended API access for Alibaba-affiliated users last month following concerns over suspected misuse.
According to the company, the suspension was linked to suspected distillation, a process in which developers use the outputs of advanced AI models to improve competing systems.
An OpenAI spokesperson told the Financial Times that the activity had been reported to the US government.
OpenAI said it does not permit access to its AI models from within China.
However, it confirmed that “some companies” with Chinese ownership or headquarters are allowed to use its tools for operations in countries where the company can enforce safeguards and monitor for distillation risks.
The company added, “We would rather see more of the world using AI shaped by democratic values than AI controlled by autocratic governments”.
Debate grows over export control policies
Although the transactions remain legal, they have prompted renewed calls for Washington to impose tighter controls on advanced AI models, similar to restrictions already placed on the export of high-end AI chips.
The US government has introduced controls covering individual frontier AI models, including Anthropic’s Mythos and Fable, as well as OpenAI’s GPT-5.6.
However, current regulations do not broadly prohibit Chinese-headquartered companies from accessing advanced AI software, including companies listed on the Pentagon’s congressionally mandated 1260H blacklist, which identifies entities alleged to have ties with the People’s Liberation Army.
Google said its AI services remain available in Hong Kong and Singapore, subject to the company’s usage policies, including restrictions against distillation.
The company said geographic sales restrictions alone are not enough to eliminate distillation risks because sophisticated users can circumvent location-based controls.
Anthropic adopts stricter restrictions
Anthropic has taken a more restrictive approach by prohibiting Chinese companies and foreign entities owned by them from using its advanced AI models.
The company acknowledged that enforcing the restrictions has been challenging.
Last week, it said it had closed loopholes that previously allowed some Chinese companies to bypass its safeguards against unauthorised access.
Anthropic has previously accused Chinese AI developers DeepSeek, Moonshot and MiniMax of engaging in distillation.
Last month, the company also told Congress that Alibaba allegedly used 25,000 fraudulent accounts to generate more than 28.8 million exchanges with Claude, claiming the activity violated its terms of service.
