US data brokers are selling half a billion dollars in RL training environments to Chinese AI labs
What happened
US data brokers including Mercor, Surge AI, and AfterQuery have sold roughly $500 million worth of reinforcement learning (RL) training environments to Chinese AI labs. These datasets include RL environments originally developed for major Western AI firms like OpenAI and Anthropic. At least nine US companies have participated in these transactions, which remain largely unaffected by current export controls.
Why it matters
This flow of advanced RL training environments to China shifts competitive dynamics in AI development. It undercuts US efforts to maintain an edge by limiting the export of sensitive AI capabilities. Chinese labs gain access to high-quality, battle-tested training environments without replicating costly development from scratch. For US companies, the lack of effective export restrictions means lucrative data sales continue despite national security concerns. This situation presses regulators to reassess what AI training data qualifies as controlled technology.
What to watch next
Expect increased scrutiny on data brokers’ roles in global AI supply chains and potential tightening of export rules around AI training environments. Investors and founders should monitor how this transfer affects AI product timelines and innovation cycles, especially in China. Watch for whether US policy moves to block or limit these sales and how Chinese labs leverage such imported resources to accelerate their RL research. The evolving export control landscape will be critical for anyone operating at the intersection of AI data, training infrastructure, and international markets.
AI Quick Briefs Editorial Desk