China continues to refine its comprehensive AI regulatory framework with a new focus on systems capable of human-like interaction. On December 27, 2025, the Cyberspace Administration of China (CAC), the country’s top internet regulator, released draft rules for public comment aimed at overseeing artificial intelligence services that simulate human personalities and foster emotional connections with users. This move builds on a series of prior regulations, including the 2023 Interim Measures for Generative AI Services and the September 2025 labeling requirements for AI-generated content, reflecting Beijing’s balanced approach to promoting innovation while mitigating risks like misinformation, psychological harm, and social manipulation.
Scope and Key Requirements The draft specifically targets “interactive artificial intelligence” designed for prolonged emotional engagement, such as companion chatbots, virtual girlfriends/boyfriends, or AI therapists. Providers must conduct security assessments, register services if they reach significant user thresholds, and implement real-time content monitoring to align with socialist core values. Outputs must be labeled as AI-generated where appropriate, and systems should avoid inducing dependency or spreading harmful ideologies. Transparency obligations include disclosing interaction mechanisms and allowing users to opt out of personalized emotional features.
Implications for Industry and Society These rules signal heightened scrutiny amid the rise of advanced large language models in China, from companies like Baidu (Ernie Bot) and Moonshot AI, which increasingly offer empathetic, personality-driven interfaces. While encouraging responsible development, the regulations could raise compliance costs and slow deployment of cutting-edge features. Public feedback is sought until late January 2026, potentially shaping final rules. Analysts view this as part of China’s proactive governance model, contrasting with more fragmented approaches elsewhere, aiming to prevent issues like AI-driven loneliness epidemics or exploitation while supporting the sector’s growth toward global leadership by 2030.