OpenAI, Google sold AI models to China's blacklisted groups

The Financial Times investigation highlights a potential gap between U.S. export control policy and the operational reality of global AI companies. Both OpenAI and Google offer AI models through web-based and API platforms that, by design, are accessible from many countries and by many users—making it technically difficult to enforce country-level or entity-level restrictions at the point of access. The fact that blacklisted Chinese organizations were able to obtain these models suggests either that the companies lack sufficient controls to screen access, or that the screening mechanisms in place proved insufficient. This tension reflects a broader challenge facing U.S. policymakers: how to maintain effective export controls over AI technology when that technology is delivered via internet-accessible services, often to end users who may have legitimate reasons to access it (researchers, developers, startups) and who may obscure their organizational affiliation. The allegation may prompt the U.S. government to clarify its expectations for AI companies operating globally, or to impose additional compliance requirements on model distribution platforms.