OpenAI and Google Sold AI Services to U.S.-Blacklisted Chinese Companies
N.R. Finch
OpenAI and Google are selling advanced AI services to Alibaba, Baidu, and Tencent — all three on the Pentagon's 1260H blacklist — through Singapore-registered subsidiaries, exposing a glaring gap in US AI export controls.
Who is buying, and how?
Alibaba, Baidu, and Tencent access OpenAI and Google AI services via Singapore-registered subsidiaries.
All three sit on the US Congress-mandated 1260H list — companies flagged for ties to the Chinese military.
This means → Washington labels these firms national-security threats while letting its own AI giants sell to them. The control chain breaks at the AI layer.
Why is this legal — and why does it matter?
Under current rules, these sales are entirely legal. The US restricts chip exports to China but has no equivalent controls on AI model services.
In plain terms = the chip door is shut, but the AI-services door is wide open — and Chinese firms walk right through it.
Chris McGuire, a tech-and-security fellow at the Council on Foreign Relations, put it bluntly: the Trump administration claims it wants to beat China on AI, "but they have done nothing on export controls."
How are the three US firms responding?
OpenAI suspended Alibaba-linked users' API access last month, citing suspected "distillation" — using a model's outputs to improve a rival system. It says it notified the US government.
Google confirms its AI services are available in Hong Kong and Singapore under its usage policies, but concedes that geographic restrictions alone cannot stop sophisticated actors.
Anthropic takes the hardest line, banning Chinese companies and their offshore entities from its advanced models — yet it plugged bypass loopholes just last week, showing enforcement lags far behind policy.
How bad is the distillation problem?
Anthropic has accused DeepSeek, Moonshot, and MiniMax of distilling its models.
Last month Anthropic told Congress that Alibaba used 25,000 fake accounts to generate over 28.8 million interactions with its Claude model, violating terms of service.
This reflects a harsh reality: even with usage agreements and account bans, industrial-scale distillation is still happening — rules on paper cannot outrun determined adversaries.
What comes next?
Alibaba filed in US court last month to be removed from the 1260H list, calling the designation "arbitrary and capricious." Baidu declined to comment; Tencent did not respond.
AI-policy expert Joe Khawam warns that if Chinese labs can systematically extract frontier capabilities without bearing the compute, engineering, and safety costs US firms shoulder, it will erode the economic foundation of America's AI lead.
Put simply = if rivals can copy the homework at a fraction of the cost, spending more does not guarantee winning. Whether the regulatory framework can catch up to reality is the central test of US AI strategy.
Content is for reference only, not financial advice.