> ## Content Index
> Fetch the complete content index at: https://adjacent.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# US data labeling firms selling training datasets to Chinese AI labs
- URL: https://adjacent.media/signals/us-data-labeling-firms-selling-training-datasets-to-chinese-ai-labs/
- Published: 2026-08-07T16:10:01.000Z
- Updated: 2026-08-07T16:10:01.000Z
- Description: Surge AI, Mercor, and other American contractors have served US government agencies and AI labs while selling identical training datasets to Chinese competitors. The leak exposes a structural gap in export controls that treat data differently than hardware.
- Author: Jonathan Greene
- Tags: #signal, theme-ai, training data, ai supply chain, geopolitical risk

Source: [Forbes](http://www.forbes.com/sites/annatong/2026/08/05/silicon-valleys-other-china-problem-its-training-their-ai/?ref=adjacent.media)

Surge AI, Mercor, and other American contractors have served US government agencies and AI labs while selling identical training datasets to Chinese competitors. The leak exposes a structural gap in export controls that treat data differently than hardware. US data labeling outsourcers operate without meaningful restrictions on geography or end-use, meaning the same human-annotated datasets used to train Claude or military AI systems are becoming inputs for Chinese model builders at lower cost. This undermines chip embargoes and model licensing—if training data flows freely across borders, hardware and software restrictions alone cannot sustain technological advantage.