/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

← → days · ↑ ↓ browse · Enter similar · o open

How data tagging firms in China are using cheap labor to apply labels to troves of images and surveillance footage to help train AI systems across the country

Li Yuan / New York Times :

New York Times Li Yuan

Context & Ripple Effects

This 2018 report was an early map of the labor layer beneath China's AI boom: domestic tagging firms applying cheap human labels to image troves and surveillance footage for the country's biggest tech customers. The years since have shown the model wasn't a local quirk but a template — by 2022, Appen's Venezuela operations showed Western labelers running the same playbook in crisis economies, and Rest of World's reporting traced how Chinese firms extended it abroad.

Two later developments sharpen why the original story matters. The same firms serving Baidu, Alibaba, and JD.com were later found pushing the work onto vocational-school internships at home, while Chinese companies hiring Kenyan annotators through WhatsApp middlemen exported the opacity along with the labor. Meanwhile, export controls pushed Chinese labs like 01.ai and DeepSeek toward smaller, more carefully chosen training sets — changing what kind of labeling the industry sells.

First-order effects

  • Baidu, Alibaba, JD.com, and China's surveillance ecosystem get trained models built on manually labeled imagery, with the cost of that supervision borne by low-wage taggers rather than the buyers.

Second-order effects

  • Labor arbitrage becomes the industry's growth engine: labeling work migrates from Chinese cities to vocational schools, Venezuelan crisis labor via Appen, and Kenyan workers recruited through opaque middleman networks, with each move lowering the price floor.

Third-order effects

  • Annotation hardens into a globalized, accountability-light supply chain spanning both Chinese and US vendors — Surge AI and Mercor among them — even as export controls push Chinese labs toward smaller curated datasets, shifting demand from bulk labeling to higher-skill curation.

The trend: AI training data is consolidating into a borderless low-wage labor market whose geography shifts with local costs and export-control pressure.