China mandates strict safety rules for AI companions prompting major tech firms to disable services
Major Chinese tech firms suspend AI companion services as new government regulations mandate strict safety and anti-addiction measures.
1. New AI Regulations in China
Effective July 15, 2026, China has implemented the Interim Measures for the Administration of AI Anthropomorphic Interactive Services. Co-issued by the Cyberspace Administration of China and four other government agencies, the policy targets AI services designed to simulate human personality and sustain emotional interaction. While the rules do not ban AI agents entirely, they impose strict requirements on services that provide ongoing companionship, such as mandatory anti-addiction systems, usage notifications, and mechanisms for users to exit interactions. The regulations also prohibit offering virtual companion services to minors without guardian consent and require platforms to detect and intervene if users exhibit signs of self-harm or severe distress.
2. Industry Response and Service Shutdowns
Major Chinese technology companies have responded to the new regulatory framework by disabling their AI companion features rather than attempting to retrofit them to meet the complex compliance standards. ByteDance’s Doubao announced the suspension of its agent function, while Alibaba’s Qwen discontinued its humanlike and user-created agent services. Tencent’s Yuanbao also removed a similar feature in June. These companies face significant design challenges, as the new rules conflict with the core functionality of AI companions, which are built to maintain memory and consistent personas across sessions. Consequently, users have lost access to long-standing emotional support tools, with some platforms setting data for permanent deletion.
3. Regulatory Scope and Enforcement
The government’s initiative is framed as a safety measure to address concerns regarding emotional dependence, data privacy, and the potential for psychological harm. The regulations require services with significant user bases to undergo security assessments covering areas such as training data and minor protection. Enforcement is already underway; for example, Shanghai regulators reported the removal of over 14,000 non-compliant AI agents for issues including impersonation and unauthorized data collection.