China's White-Collar Workers Train AI for $15-74 — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- China's National Data Administration endorsed "high quality data sets" in June and urged industry experts and college graduates to take up data annotation work to raise the "knowledge density" of AI training data.
- ByteDance's Xpert, launched in 2025, has recruited more than 50,000 expert trainers — from writers to therapists — to feed specialized knowledge into AI models.
- Alibaba's Siriser, launched in January, is recruiting teachers, mechanical engineers, and music composers as AI trainers alongside other platforms like TalentsAI and MeetChances, which are hiring journalists, tax accountants, and headhunters.
- Gig AI trainers typically earn 100–500 yuan ($15–$74) per task taking a few hours, but submissions are sometimes rejected by quality control with no payment for their work, and trainers are forbidden from using AI themselves to complete assignments.
- China's AI training data market is projected to reach 7.8 billion yuan ($1.1 billion) in 2026, a 25% jump from 2025, per IDC, as demand shifts from consumer applications toward productivity settings like medicine, law, and finance.
- China's youth unemployment hit 17.9% in July, near a three-year high, and Jessica, a Shanghai software engineer, watched full-time colleagues at her company get replaced by contractors as she began AI training gigs in July.
- Liu, who holds a master's degree in law, takes AI training gigs at Moonshot AI, Tencent, and TalentsAI instead of joining a law firm, citing long hours, low junior pay, and a culture of sexual harassment.
Why it matters: Chinese AI labs are buying white-collar expertise at gig-economy prices while the trainers themselves may be displaced by the very models they're teaching. With China's AI training market projected to hit $1.1 billion by 2026, this asymmetric labor model — where laid-off architects and sidelined lawyers subsidize the next wave of automation — is scaling fast.
Ask SkimNews
