OpenAI and Google sold AI services to Singapore-based subsidiaries of Alibaba, Baidu, and Tencent, reigniting calls for tighter US regulation of AI models
US groups have been supplying AI services to Singapore-based subsidiaries of Alibaba, Baidu and Tencent.
Context & Ripple Effects
Earlier coverage described OpenAI cutting off API access to Chinese developers, while Chinese companies including Alibaba, Baidu and DeepSeek pursued open-source models as a way to reduce exposure to US curbs and draw in outside development. The reported use of Singapore-based affiliates shows that access controls aimed at developers or entities can be harder to apply across multinational corporate structures.
The story also lands as Chinese authorities have discussed limiting foreign access to advanced models and penalizing AI theft. That creates a two-sided policy contest: US restrictions seek to constrain access, while Chinese policy is considering how to govern reliance on foreign models and domestic model development.
First-order effects
- OpenAI and Google face renewed scrutiny over customer screening and the extent to which service provision through overseas subsidiaries complies with the intent of US AI controls.
- Alibaba, Baidu and Tencent's Singapore-based units retain a channel to advanced US AI services, but that channel becomes more politically and regulatory exposed.
Second-order effects
- Calls for tighter rules could push US model providers and cloud partners to apply more granular affiliate, beneficial-ownership and usage checks, raising compliance friction for multinational customers.
- Chinese AI groups have another incentive to diversify away from foreign model access through domestic and open-source alternatives, reinforcing the strategy already visible among companies such as Alibaba, Baidu and DeepSeek.
Third-order effects
- If policymakers treat overseas affiliates as a persistent loophole, AI governance is likely to move beyond country-based access rules toward controls focused on corporate control, end use and service delivery—not just where an account is registered.
- The episode points to a more segmented AI stack, in which leading model access, open-source distribution and domestic approvals become parallel routes for companies navigating US-China technology restrictions.
The trend: This is one data point in the shift from broad, geography-based AI restrictions toward contested controls over who can access frontier models through global corporate and cloud structures.