China Cracks Down on AI Abuse and 'Digital Slop' Across Major Models
China's Cyberspace Administration (CAC) announced the second-phase results of its campaign against AI misuse, removing over 5.61 million illegal posts, penalizing 49,000 accounts, and sanctioning 2,400 apps and websites. Leading domestic AI models—including Doubao, Yuanbao, Qwen, and Ernie Bot—are now subject to strict raw data auditing and mandatory AI-content labeling rules. This regulatory enforcement highlights China's proactive approach to AI governance, holding both foundation model providers and social media platforms directly responsible for synthetic media misuses. It sets a strict compliance precedent that mandates end-to-end control over deepfakes, low-quality automated content, and algorithmic manipulation. Targeted violations include deepfake celebrity impersonations, automated bot net services, 'toxic animations' aimed at minors, and low-quality AI rewrites of classic literature dubbed 'digital slop.' Major distribution platforms upgraded their multimodal detection systems and dynamically expanded faceprint and voiceprint databases for real-time content interception.
## BACKGROUND
China has implemented mandatory national standards requiring clear labeling and digital watermarking on all AI-generated content to help users distinguish real from synthetic media. Concurrently, the rise of 'digital slop'—low-effort AI content optimized specifically to exploit recommendation algorithms—has flooded social feeds and increasingly marginalized human creators.