US Data Brokers Sell $500 Million in RL Training Environments to Chinese AI Labs

At least nine US data firms are selling roughly $500 million in AI training data to Chinese AI labs, according to an investigation by ChinaTalk. Some of that data consists of reinforcement learning environments, simulated worlds used to train AI models on tasks such as coding, financial analysis, and office work. The report says Chinese labs buy the same products previously sold to US frontier labs, allowing them to reduce costly development and training runs.

Who is involved

American data firms named in the report include Mercor, Surge AI, AfterQuery, and Turing. AfterQuery reportedly derives a significant share of its annual recurring revenue from Chinese customers, and ChinaTalk cites Forbes reports of about $100 million in ARR while saying it could not independently verify the sales. Surge AI openly advertises on its website that its products are built by the same experts behind its frontier-lab work. The deals also run through offshore structures, with ByteDance maintaining procurement offices in Singapore and San Jose. The relationships came to light through anonymous sources and sloppy operational security among subcontractors, according to ChinaTalk. The US data firms outsource much of the RL environment creation and data curation to low-paid workers in developing countries, especially India, and those workers sometimes post client names on LinkedIn or in resumes despite being expected to use code names.

How Chinese labs benefit

Chinese labs deliberately purchase environments already sold to Anthropic or OpenAI, so they know the data works before using it. That helps them avoid in-house development and wasted training runs, which is especially valuable because Chinese chips face acute memory bottlenecks. The report also says Chinese labs use powerful US models such as GPT or Claude as teachers, using their outputs as training examples. Together, purchased RL environments and distillation create a two-pronged advantage that lets Chinese labs close the gap with less compute of their own.

Dual-use concerns

The trade remains largely unregulated, and the authors argue that the Bureau of Industry and Security should subject dual-use environments to existing export controls. Coding environments can improve software development but may also sharpen offensive cyber capabilities. Biology environments could improve gene synthesis capabilities and potentially enhance bioweapons-relevant applications. The authors also stress that the relationships are not apocalyptic: by piggybacking on US work, Chinese labs still trail their US peers by at least a few months as a structural matter, and sensitive environments in biology, chemistry, cybersecurity, chip design, and investment banking are increasingly built in-house at American frontier companies.

Sources: The Decoder