Cuicui, a mid-40s architect based in Shenzhen, has spent years designing public spaces like stadiums, schools, and hospitals. However, government spending on infrastructure has declined over the past five years, cutting her income in half. To manage her mortgage and her child’s extracurricular expenses, she now works as a specialized AI trainer. After her regular job, she logs on to a platform called TalentsAI, where she provides AI models with detailed instructions on tasks like writing building proposals or conducting financial analyses. "It’s like teaching a child," she said. "Every task has to come from my actual work." In the U.S., companies like Mercor and Surge AI have already hired tens of thousands of professionals—filmmakers, lawyers, and scientists—to train AI systems in complex tasks. Now, a similar trend is emerging in China, where the government is pushing for widespread AI adoption. In June, China's National Data Administration encouraged the creation of "high quality data sets" and urged more industry experts to contribute to AI training. This has led to an increase in data annotation work, with many recent college graduates taking on flexible, part-time roles in the field. Freelance AI trainers, like Cuicui, often request anonymity to avoid being identified by employers or platforms. They describe the work as unstable, demanding, and lower paid compared to similar roles in the U.S. However, with a tough job market and the rise of AI, many professionals have taken on these tasks to earn income and stay involved in what they see as the future of industry. "What can middle-aged people do?" one trainer asked, reflecting the uncertainty many face. Alibaba and ByteDance are two of China's leading tech giants investing heavily in AI training. Alibaba's platform, Siriser, recruits professionals like teachers and engineers, while ByteDance's Xpert has hired over 50,000 experts, from writers to therapists. Independent platforms like TalentsAI and MeetChances also hire journalists, tax accountants, and headhunters to train AI systems. On social media, platforms like Xiaohongshu have posted messages encouraging middle-aged workers to consider AI training as a new career path. Chinese AI labs are now on par with Silicon Valley in developing models that can perform complex reasoning and coding tasks. Companies are focusing more on specialized data collection to improve AI tools for professional use. Tech giants like Tencent and ByteDance have launched AI office tools that can generate presentations and websites. However, training AI for fields like medicine and finance requires more detailed data. "Making models perform better at work will be a strong differentiator in future AI competition," said Tiezhen Wang, a researcher based in Sydney. Despite the challenges, many data workers see AI training as a way to diversify their income in an uncertain economy. With youth unemployment reaching 17.9% in July, many graduates are struggling to find work. Liu, a Shenzhen-based AI trainer with a law degree, chose to avoid traditional legal jobs due to long hours and a toxic work culture. She now trains AI models to provide legal advice and draft judgments. She plans to take the civil service exam for more stability. Over the past decade, China's data annotation industry has primarily relied on low-paid workers, including students and rural residents, to support the development of self-driving software and voice recognition tools. However, demand is now shifting toward productivity-focused AI applications. According to IDC, China's AI training data market is expected to grow significantly in the coming years. Julie Yujie Chen, a researcher at the University of Toronto, noted that middle-class professionals are increasingly entering the field due to economic pressures and government support. Jessica, a Shanghai-based software engineer with two decades of experience, has started training AI models to test their ability to process complex texts and software skills. While the pay is low, she sees it as a way to stay relevant in an industry rapidly transformed by AI. "You never know how long you will keep your job," she said. "I want to have one more option. You have to keep evolving yourself to avoid getting completely replaced."