ByteDance and Alibaba disable custom AI agents ahead of new regulations

The immediate suspension of consumer-facing custom agents marks a shift toward productivity-centric AI and away from unregulated companionship models. For the streaming and digital interaction ecosystem, this indicates that personalized, high-engagement synthetic personas face a significantly higher regulatory bar than specialized task-based tools. The move forces major players to pivot their consumer offerings toward functional utility as Beijing codifies identity and traceability standards. Market participants should monitor whether these restrictions lead to a permanent fracturing between industrial 'productivity' AI and consumer socio-emotional AI, potentially cooling the market for virtual companion services.

The rollout of the Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services, effective July 15, 2026, represents the world’s first specialized legislation targeting 'emotional AI.' Per MMLC Group (April 2026), these rules mandate a closed-loop management system that includes algorithm filing, security assessments, and specific prohibitions against providing virtual relatives or intimate partners to minors. The regulation also bars using sensitive user conversation data to train future models, according to reporting from Business Times (July 2026).

This move aligns with broader state efforts to formalize the AI infrastructure. On June 26, 2026, the State Administration for Market Regulation (SAMR) released seven national standards for AI-agent interconnection. Per China Daily, these standards establish a unified framework for digital identity, interaction protocols, and tool invocation, involving over 70 companies including Xiaomi and Kuaishou. The goal is to create a standardized 'identity code' system for agents, facilitating secure cross-platform interaction while ensuring every agentic action is traceable to a registered entity.

While curbing emotional AI, Beijing is aggressively promoting industrial adoption. Per City News Service (May 2026), a separate regulatory blueprint set a goal for a 70% adoption rate of intelligent agents across major industries by 2027. This framework encourages applications in manufacturing, financial risk control, and scientific research. The divergence in policy—tightly restricting social companionship while subsidizing industrial deployment—suggests that China views AI agents as critical productivity infrastructure that must be decoupled from the potential psychological and social risks identified in consumer-facing anthropomorphic bots.