In the lead-up to China’s 2025 International Workers’ Day, major Chinese tech firms escalated their artificial intelligence race, with Alibaba releasing the Qwen 3 large language model (LLM), Xiaomi entering the market with Mimo, and DeepSeek unveiling the Prover V2 model. Strategically timed product launches positioned these models ahead of the holiday, but the celebration of worker solidarity has become largely symbolic for the nation’s most vulnerable, including over 300 million rural migrant workers. Notably, Chinese-developed artificial intelligence models exhibit heavy censorship by suppressing information about ongoing social movements, especially labor protests, raising urgent questions about their impact on workers´ rights and visibility.
Throughout early 2025, nearly 540 worker protests in China were documented by the China Labor Bulletin. However, when users asked DeepSeek to list labor protests, the system deflected with official narratives about ‘social harmony’ and government protections, denying or omitting protest activities even when presented with specific incidents. The censorship intensified in Mandarin interactions, with warnings against spreading ‘false information’. Alibaba’s Qwen 3 censored discussions labeled as ‘disturbing social order’ and was selective in addressing labor issues, providing more open responses in English and more conservative ones in Chinese, particularly on sensitive topics such as forced labor in Xinjiang and Tibet. Developers can rapidly update these models to align with party directives, heightening suppression of critical content.
These models are further contributing to a sophisticated surveillance ecosystem. DeepSeek has been integrated into state social monitoring systems, such as GoLaxy’s Comprehensive Social Listening System, enabling heightened scrutiny of activities on both public and semi-private platforms popular among working-class citizens. DeepSeek’s adaptation for police use increases risks for labor activists and dissidents, with documented instances of coordinated crackdowns. The integration of artificial intelligence into petition systems—used by workers to lodge grievances—raises concerns about intensified persecution and surveillance of workers engaged in labor disputes.
The transformation extends to workplace realities. Sectors like food delivery, dominated by platforms such as Meituan, have implemented algorithm-driven systems that subject workers to unsafe conditions, strict deadlines, and little recourse, reflecting both new forms of algorithmic oppression and the longstanding lack of worker bargaining power. Efforts by worker advocates such as Cheng Guojiang have been met with arrests. Similarly, the data labeling industry underpinning China’s artificial intelligence boom is characterized by poor working conditions, long hours, and wage opacity, deepening social and economic divides. In 2025, the International Trade Union Confederation continues to rank China as a nation with no effective rights for workers, underscoring how artificial intelligence advancement, as currently practiced, risks deepening labor exploitation and eroding collective worker agency.
