Data Marketplaces & Intermediaries

US data-labeling firms reportedly sell training data to Chinese AI labs, including Tencent

Source: AI Weekly · Aug 11, 2026

This is the most structurally uncomfortable story in this batch, and the one most directly relevant to a data-provenance-focused business like ours. The core allegation: several of the same US data-labeling companies that serve American AI labs — and in some cases the US government and military directly — are reportedly, simultaneously, selling training datasets to Chinese AI labs, several of which carry real geopolitical weight.

The Chinese buyers named include Tencent, which the Pentagon has formally designated a Chinese military company, alongside ByteDance, Alibaba, Ant Group, and Moonshot. On the American side, Surge AI is reported to hold verified contracts with the US Army and Air Force, while separately generating $1.2 billion in 2024 revenue from customers including Google and Anthropic — meaning the same company's client roster reportedly spans US military, frontier AI labs, and (per the allegation) Chinese AI labs simultaneously. AfterQuery is reported to earn more than $50 million annually specifically from Chinese lab customers. Mercor's reported Chinese-source revenue is proportionally smaller — around 2% of Q2 revenue — but with the company's overall revenue scaling as fast as it is, that percentage still represents meaningful and growing dollar terms. In aggregate, the reporting puts the top six Chinese labs' combined annual spend with American data-labeling vendors at roughly $500 million.

What's conspicuously absent from the reporting is any response from the Chinese companies themselves — ByteDance, Alibaba, Moonshot, Tencent, and Ant Group reportedly did not respond to press inquiries — and equally notably, the vendors' own explanation of how they think about this dual-market exposure isn't represented either. It's also genuinely unclear from the reporting whether these arrangements cross any formal export-control line, or whether they sit in a gray zone of unwritten policy expectations that companies are choosing to interpret generously.

For a business built on data provenance and clean chain-of-custody as a core value proposition, this story is close to a direct validation of the market gap we're positioned against. If a customer or partner can't get a straight answer about where a vendor's data pipeline also flows, provenance guarantees become the differentiator, not a compliance checkbox. Worth watching closely for how (or whether) this develops into formal regulatory scrutiny.

Key Points

• Reported buyers include Tencent (Pentagon-designated Chinese military company), ByteDance, Alibaba, Ant Group, and Moonshot
• Surge AI's verified US clients reportedly include the Army and Air Force, alongside $1.2B 2024 revenue from Google and Anthropic
• AfterQuery reportedly earns $50M+/year from Chinese labs; Mercor's Chinese-source revenue is smaller (~2% of Q2) but scaling with overall growth
• Aggregate: the top six Chinese labs reportedly spend ~$500M/year with American data-labeling vendors
• None of the named Chinese companies responded to reporter questions; whether this violates export controls or only unwritten policy lines is unclear