OpenAI and Google Provide AI Models to Chinese Companies on Pentagon Blacklist

OpenAI and Google have provided access to AI models to Chinese companies on the U.S. Department of Defense's Section 1260H blacklist, which identifies entities linked to the People's Liberation Army. The incident involves subsidiaries of Alibaba, Baidu, and Tencent in Singapore that are listed but the list itself does not automatically trigger sanctions.

Facts Restored

Around July 10, 2026, OpenAI and Google confirmed providing advanced AI model access to the aforementioned Singapore subsidiaries. OpenAI explicitly stated that models are not allowed to be accessed directly from China's mainland, but permitted certain Chinese-controlled enterprises to use its tools in foreign countries where enforceable security safeguards are in place. Google confirmed that its advanced AI services are fully available in international hubs such as Hong Kong and Singapore, provided users comply with strict usage policies. Last month, OpenAI suspended access for several Alibaba-linked users over prohibited "distillation" operations, where outputs from a competitor's advanced AI are fed into one's own lower-tier models for rapid training.

Mechanism Breakdown

Cloud API distribution is the core operational method. AI models are delivered via commercial APIs, partners, or intermediary services, differing from traditional defense technology reliant on physical supply chains, making complete restriction difficult once models are commercialized. OpenAI, within the 24-hour window of the global deployment of the GPT-5.6 series, simultaneously expanded three models—Sol, Terra, and Luna—distinguished by intelligence, speed, and pricing, covering ChatGPT, Codex, and APIs. An OpenAI spokesperson expressed a preference for seeing more of the world using AI shaped by democratic values rather than AI controlled by authoritarian governments, and argued that nationality alone should not determine access rights. Google adopted a similar stance, emphasizing that geographic sales restrictions are no longer sufficient to fully prevent distillation risks.

Industry Impact

For developers, API access stability faces new uncertainties. OpenAI's suspension of Alibaba-linked users demonstrates its monitoring system can identify distillation behavior, but also means developers need additional verification of their compliance pathways to avoid misjudgment due to affiliations. For enterprise users, especially overseas entities of multinational Chinese tech companies, service continuity depends on proving that the operating location meets OpenAI and Google's security enforcement requirements. For the competitive landscape, upstream chip export controls have already tightened, and this incident may push Congress to require mandatory KYC for API access or stricter entity bans on adversary countries, directly impacting business models reliant on global API revenue.

Strategic Assessment

Based on existing facts, the most likely development is that regulators will require AI companies to disclose more KYC data and distillation detection logs to verify that models are not indirectly accessed by blacklisted entities. Signals to watch include whether the U.S. Department of Commerce expands export control lists to cloud service providers, and whether OpenAI further tightens access policies in hubs like Singapore.

Compared to historical precedents, previous chip export controls have been repeatedly expanded. This case, where AI models indirectly enter blacklisted entities through Singapore subsidiaries, reveals that the vulnerabilities in software distribution channels differ from hardware control logic. Both OpenAI and Google position themselves as key partners in U.S. AI dominance yet simultaneously provide access to listed entities, exposing the tension between commercial expansion and national security positioning.