OpenAI and Google have been providing AI model access and API services to Singapore-incorporated subsidiaries of major Chinese technology companies including Alibaba, Baidu, and Tencent, according to a Financial Times investigation published on July 10.
The practice raises questions about whether the arrangement complies with U.S. export control regimes designed to restrict Chinese access to advanced American AI technology. Singapore-based subsidiaries of Chinese firms operate under different jurisdictional rules than their mainland parent companies, creating a potential gap in enforcement.
The Compliance Question
U.S. export controls on AI technology have expanded significantly since 2022, with the Bureau of Industry and Security tightening rules on advanced chips and, more recently, on AI model weights and API access. The controls target entities that are owned by or affiliated with companies in countries of concern, but enforcement relies on knowing the ultimate beneficial ownership and end-use of the technology.
The FT’s findings, according to their investigation, suggest that Singapore-incorporated entities owned by Chinese tech conglomerates have been able to access American AI models through standard commercial channels. Neither OpenAI nor Google has publicly addressed the specific arrangements described in the FT report.
The Distribution Tension
The revelation exposes a structural tension in the AI industry. American AI companies face pressure to grow revenue and expand their global footprint, while U.S. national security policy increasingly treats advanced AI capabilities as strategically sensitive exports. The Singapore subsidiary structure, as described by the Financial Times, represents one way commercial distribution has flowed around geographic restrictions.
For the agent ecosystem, the implications are twofold. Agent platforms that build on OpenAI or Google models may need to implement their own know-your-customer checks on downstream API consumers, as regulatory scrutiny of indirect model access intensifies. And if the Commerce Department tightens rules in response to these revelations, it could restrict how and where cloud-hosted AI agents can operate globally.