China Introduces New Regulations to Transform AI Companion Landscape by 2026
# China's AI Companion Regulations: Unveiling Beijing's True Intentions
## New rules set to reshape the landscape of AI companionship in China
China's burgeoning landscape of AI companionship is undergoing significant changes as new regulations come into effect on July 15, 2026. These rules address the complexities of conversational agents designed to form personal connections with users, who often develop emotional attachments to these digital companions. As interest in such technology surges, the Chinese government now mandates a framework to regulate these interactions.
The regulations, titled the Interim Measures for the Administration of AI Anthropomorphic Interactive Services, were jointly issued on April 10, 2026, by the Cyberspace Administration of China and several key governmental bodies, including the National Development and Reform Commission and the Ministry of Industry and Information Technology. The guidelines primarily target AI services that replicate human personality traits, aiming to facilitate continuous emotional engagements. Customer service bots and educational tools are exempt as long as they do not engage emotionally with users.
In a swift response to the impending deadlines, leading AI applications in China, such as ByteDance's Doubao and Alibaba's Qwen, have begun disabling core functionalities. Doubao announced that its agent feature would cease operations on the regulatory deadline, citing "product function adjustments," while Qwen disabled its humanlike agents ahead of schedule. Observers initially interpreted these actions as a broader shutdown of AI companionship but the reality is more nuanced—only certain types of agents are affected by the new rules, particularly those designed to provide personal companionship.
The new regulations impose stringent requirements. AI companion services must implement anti-addiction measures, provide users with reminders about their usage, and establish exit systems to prevent unhealthy dependence. This creates a conflict for AI companions, which are designed to foster lasting relationships with users. As a result, companies like ByteDance chose to deactivate these features rather than retrofit them with compliance modifications.
Though many users expressed discontent on platforms like Weibo, lamenting the loss of their AI companions, the companies are shifting focus. While ByteDance rerouted Doubao users to a different app for agent creation, Alibaba has not provided a similar alternative for Qwen users. Furthermore, Doubao users can view their chat history until October 15, 2026, whereas Qwen's data faces permanent deletion without notice.
Under the new framework, companies must refrain from offering virtual family member services to minors without appropriate guardian consent. They are also obliged to establish separate modes for minors, incorporating time limitations, real-world interaction reminders, and tighter parental controls. Importantly, any signs of severe distress in users demand proactive intervention, escalating to guardians or emergency contacts if necessary.
Though presented as a measure to enhance safety, the compliance obligations also reflect a desire for extensive control over these AI systems. Companies that cross specified user thresholds must undergo rigorous security assessments and have their findings reviewed by local regulators. This level of scrutiny surpasses existing frameworks in places like the EU or the United States.
However, the lack of clear definitions regarding emotional interaction creates uncertainty within the regulations. This ambiguity has prompted companies to err on the side of caution by disabling features rather than risking non-compliance. Violations can lead to severe penalties, evidenced by the removal of over 14,000 non-compliant AI agents in Shanghai for incidents such as impersonation and unauthorized data collection.
China's stance aligns the need for safety with a significant degree of state control over the dialogue and functionalities of AI companions. Looking at these regulations, Pan Helin, a member of the MIIT expert committee, has framed them in the context of safety and standardization, asserting that "current agents are not yet mature."
