OpenAI and Google allegedly circumvented US export controls by supplying AI models to blacklisted Chinese companies via Singapore.

OpenAI and Google have been selling advanced artificial intelligence models to Singapore-based subsidiaries of Chinese companies that appear on a Pentagon blacklist, according to reporting by the Financial Times and confirmations from both tech giants. The revelation exposes a significant gap in Washington's efforts to restrict China's access to cutting-edge AI technology.

The parent companies involved—Alibaba, Baidu, and Tencent—are three major Chinese technology firms with connections listed by the U.S. Department of Defense as potentially linked to China's military.

While these transactions did not violate U.S. law, they have reignited debate about whether Washington's approach prioritizes semiconductor restrictions at the expense of regulating frontier AI software, which remains subject to weaker export controls. Commercial access to these AI services through cloud-based application programming interfaces (APIs) is lightly regulated, creating a compliance gap.

The Chinese companies accessed these AI services through Singapore-registered subsidiaries, arrangements that fell within existing legal frameworks. Security experts warn, however, that such structures can still grant advanced AI capabilities to organizations the U.S. seeks to restrict. Of particular concern is "model distillation"—a process in which outputs from advanced AI systems are used to train competing models.

OpenAI has since suspended API access for certain Alibaba-affiliated users after identifying potential misuse concerns. Google has similarly acknowledged the threat, admitting that geographic sales restrictions alone are insufficient to prevent model distillation risks. The controversy underscores a fundamental shift in how AI models are viewed: increasingly as strategic assets rather than merely software products.

The disclosures come as both Washington and Beijing move to tighten control over advanced AI technologies. The U.S. has intensified scrutiny of frontier AI models alongside semiconductor exports, while China reportedly plans to limit foreign access to its own advanced AI systems—marking the beginning of a period of reciprocal technology controls between the two superpowers.

Approaches within the American AI industry remain inconsistent. While OpenAI and Google have permitted limited overseas access under compliance safeguards, Anthropic has adopted a stricter policy, blocking all Chinese entities from accessing its frontier models due to security concerns. This episode raises a fundamental question for policymakers: if AI models increasingly determine economic competitiveness and national security, should governments regulate software exports with the same rigor applied to semiconductor shipments?