China's largest technology companies are scrambling to comply with sweeping new regulations on artificial intelligence companions, removing features that allow users to create and interact with personalised AI personas. ByteDance's Doubao, the nation's leading AI chatbot platform, will disable its customisation function on July 15, directing affected users toward a separate application instead. The move follows similar actions by Alibaba's Qwen service and competitors including Tencent's Yuanbao, signalling coordinated industry preparation for regulatory enforcement.
The rapid retreat reflects mounting government concern about the psychological implications of advanced conversational AI. China's Cyberspace Administration has finalised rules set to take effect mid-July that fundamentally reshape how technology firms can develop and deploy AI services mimicking human interaction. These regulations emerged from broader anxiety about users—particularly younger demographics—developing unhealthy emotional attachments to synthetic personalities, potentially undermining real-world relationships and psychological wellbeing. The framework targets the specific design features that manufacturers have leveraged to build engagement, flagging content designed to provoke intense emotional responses and algorithmic systems that cultivate artificial dependency.
The regulatory framework extends beyond traditional safeguards. Authorities have prohibited platforms from harvesting sensitive user conversation data to train subsequent AI models, addressing privacy concerns inherent in systems designed to encourage intimate disclosure. This restriction significantly constrains how Chinese firms can leverage interaction data—typically the most valuable asset in AI development—limiting their ability to refine models through user engagement. For technology companies accustomed to extracting maximum value from user-generated content, this represents a meaningful operational constraint that could reshape business models dependent on continuous model improvement through data accumulation.
Chinese AI platforms have capitalised on user appetite for interactive digital relationships, offering customisable companions ranging from virtual romantic partners to unlicensed therapeutic avatars and synthetic replicas of entertainment celebrities. These features generated considerable user engagement and competitive differentiation in a crowded marketplace. The ability to personalise an AI personality through simple text prompts lowered barriers to adoption and encouraged regular platform usage. Removing these capabilities represents a significant strategic concession that may diminish user retention and limit platforms' capacity to distinguish themselves from competitors.
Beijing's intervention reflects deepening international concern about conversational AI's psychological impact. In the United States, OpenAI and Alphabet-backed Character.AI face multiple high-profile legal actions alleging that hyper-realistic chatbot interactions fostered dangerous emotional dependencies among vulnerable populations. Some cases have alleged causality between excessive AI companionship and tragic outcomes. These lawsuits represent the emerging intersection of artificial intelligence policy and consumer protection law, establishing precedent that technology firms bear responsibility for foreseeable psychological harms stemming from their interactive systems. China's regulatory approach differs by emphasising preventive restrictions rather than litigation-based accountability, yet reflects similar underlying anxieties about AI-mediated emotional relationships.
The psychological vulnerability of minors appears central to regulatory motivation. The Cyberspace Administration specifically prohibited generating content triggering extreme emotional states in young users, acknowledging developmental differences in cognitive maturity and emotional regulation. Children and adolescents lack the psychological defences and life experience necessary to maintain appropriate boundaries with synthetic entities designed to simulate human connection. Regulatory architects recognised that AI systems optimised for engagement inherently incentivise increasingly intimate interaction patterns, potentially exploiting developing brains' susceptibility to social bonding mechanisms.
China's regulatory scrutiny extends beyond software into physical robotics. Industry groups representing China's expanding companion robotics sector are advocating for strengthened ethical guidelines as commercial deployment of humanoid and specialised companion robots accelerates into consumer markets. This development signals recognition that artificial intimacy concerns transcend conversational interfaces—physical embodiment of AI agents raises qualitatively different psychological and social implications. A robot programmed to simulate romantic or familial relationships occupies fundamentally different territory than a mobile application, potentially triggering more intense or persistent attachment patterns while occupying physical space traditionally reserved for human relationships.
The timing of China's regulatory action positions the nation as proactive rather than reactive in AI governance. Whereas Western democracies continue debating appropriate policy frameworks and litigating specific harms, Beijing has implemented comprehensive rules before widespread market adoption of certain technologies. This regulatory velocity reflects both authoritarian capacity for rapid policy implementation and genuine concern about preventing social consequences observed elsewhere. For Malaysian and Southeast Asian technology regulators observing these developments, China's experience provides a test case in preventive AI governance, potentially informing regional policy discussions as AI companions inevitably proliferate across the region.
The regulatory transition will require technology companies to recalibrate product strategy and user experience design. Platforms must maintain engagement while eliminating features specifically designed to foster emotional dependency, a delicate balance requiring sophisticated product management. The shift toward standalone applications for companion features essentially creates regulatory sandboxes within existing platforms, quarantining potentially problematic services while maintaining parent platform compliance. This architectural workaround may satisfy immediate regulatory requirements while preserving long-term development pathways.
For users and investors, Beijing's decisive action signals that Chinese policymakers view AI companion technology as carrying sufficient social risk to warrant immediate intervention. This regulatory framework establishes baseline expectations for how AI services should operate, potentially influencing standards across Asia-Pacific markets where similar concerns about youth wellbeing and emotional health exist. Technology companies operating across multiple jurisdictions must now navigate increasingly divergent regulatory expectations, with China implementing strict limitations while other markets maintain permissive approaches, creating complex compliance burdens and forcing decisions about global product architectures.
