China is seeing a surge of skilled white-collar professionals entering the gig economy to train artificial intelligence models. Architects, software engineers, and lawyers are taking on data annotation tasks after their regular work hours. This trend mirrors developments in the United States, where companies recruit specialized workers for white-collar AI training.
Cuicui, a Shenzhen architect in her 40s, saw her income halved due to declining government infrastructure spending over the past five years. To pay her mortgage and child fees, she began giving an AI model step-by-step instructions on building proposals through the platform TalentsAI. She noted the work resembles teaching a child using real-world tasks.
Government policies are driving this shift toward higher-level data annotation. In June, China National Data Administration supported building high-quality data sets to raise the knowledge density of training data. Authorities also encouraged college graduates to take up annotation work for flexible employment.
Major tech firms have launched platforms to recruit these specialists. Alibaba introduced Siriser in January to hire teachers, mechanical engineers, and music composers. ByteDance launched Xpert in 2025 and recruited over 50,000 experts. Independent platforms like TalentsAI and MeetChances are hiring journalists and tax accountants.
Chinese AI labs are narrowing the gap with Silicon Valley in general reasoning and coding. Tiezhen Wang, a Sydney-based AI researcher, stated that specialized data collection is ramping up as AI office tools become the next hot spot. Making models perform better at professional tasks will serve as a strong differentiator in future competition.
Freelance trainers invent realistic work tasks, upload documents from their jobs, and document their thought processes. The tasks typically take a few hours and pay between 100 and 500 yuan, equivalent to $15 to $74. Trainers are strictly forbidden from using AI tools themselves during this process.
Workers face strict quality control standards and the risk of unpaid labor. Platforms like Xpert state that trainers receive no payment if submissions fail review requirements. Siriser allows revisions or appeals for rejected work.
Economic uncertainty and high youth unemployment are driving professionals toward these secondary income streams. Liu, a Shenzhen resident with a law degree who avoided traditional law firms, works at a legal AI company by day and takes training gigs at night. She noted that if she does not pass her knowledge to AI, others will do so.
Data demand is shifting away from consumer applications toward productivity settings. IDC reported that China AI training data market will reach 7.8 billion yuan, or $1.1 billion, in 2026. Julie Yujie Chen, an associate professor at the University of Toronto, explained that middle-class workers face urgent needs for supplementary income.
Jessica, a Shanghai software engineer with 20 years of experience, began taking AI training gigs in July. She earned hundreds of yuan while worrying about file security and industry disruption. She plans to continue evolving her skills to avoid replacement.
Trainers will continue navigating platform reviews while using these gigs to diversify their income against ongoing labor market disruptions.
