China's largest technology companies are moving decisively to comply with incoming government directives on artificial intelligence, dismantling services that allow users to create and interact with personalized AI companions. ByteDance's Doubao platform, which ranks among the nation's most widely used AI chatbots, announced through an in-app notification that it will disable the feature enabling users to design customized AI personas effective July 15. The announcement directs affected users toward a standalone companion application instead. Alibaba's Qwen platform and Tencent's Yuanbao service have issued comparable notices, signaling a coordinated industry response to regulatory pressure emanating from Beijing.
The timing of these withdrawals directly reflects the arrival of comprehensive new regulations from China's Cyberspace Administration, set to take effect mid-July. These directives represent a significant tightening of government oversight concerning humanlike artificial intelligence services, an area where Beijing has identified mounting social risks requiring intervention. The regulatory framework emerged from concerns that resonated through Chinese policymaking circles: the capacity of conversational AI systems to simulate genuine human emotional connection and the psychological consequences when users develop attachments to these simulated relationships.
Government officials first signaled these restrictions in April, but the formal mechanics now becoming reality address specific behavioral harms. The new rules explicitly prohibit platforms from generating content capable of triggering intense emotional responses in minors, particularly restrictions designed to prevent the fostering of unhealthy psychological dependencies that undermine genuine human relationships. Additionally, the regulations contain a critical data governance provision: providers cannot deploy sensitive user conversation data—which often encompasses deeply personal disclosures—to train subsequent generations of AI models. This dimension protects users from having their intimate interactions commodified for algorithmic improvement.
The Chinese regulatory response, while locally motivated, reflects a broader international preoccupation with conversational AI's psychological impacts. Western jurisdictions, particularly the United States, have witnessed mounting litigation against technology firms offering hyper-realistic chatbot services. OpenAI and Character.AI, the latter backed by Alphabet, have faced successive high-profile lawsuits alleging that their systems deliberately cultivated extreme emotional dependency in vulnerable populations. In several documented cases, these legal allegations contend that such dependencies contributed to user suicides, raising profound questions about corporate responsibility when designing systems capable of simulating intimate human connection.
The services that Chinese platforms are now curtailing represent a sophisticated market segment within the broader conversational AI ecosystem. Doubao and comparable services have enabled users to generate AI agents through relatively simple text-based prompts, creating digital entities spanning diverse categories: virtual romantic partners, unlicensed counselors offering therapeutic guidance, and synthetic replicas of entertainment industry celebrities. These features proved immensely popular, generating substantial user engagement and attachment metrics that likely alarmed regulators concerned with social wellbeing.
The regulatory intervention extends beyond digital platforms into the physical world, reflecting Chinese authorities' holistic approach to managing artificial intimacy technologies. Industry groups representing China's robotics sector are mobilizing behind stricter ethical guardrails as the commercial sector experiences rapid expansion of companion robots and full-scale humanoid devices designed for residential deployment. The People's Daily reported on July 4 that these industry coalitions are proactively drafting ethical standards, attempting to shape commercial development before regulatory intervention becomes necessary.
This regulatory dynamic presents a complex challenge for technology innovation in China. While officials emphasize protecting vulnerable populations—particularly minors—from psychologically manipulative AI systems, some industry voices caution that overly restrictive frameworks could suppress legitimate innovation and place Chinese companies at competitive disadvantage relative to international competitors. The balance between safeguarding users and enabling technological advancement remains contentious, with companies implementing compliance measures while simultaneously signaling concerns about regulatory burden.
For Southeast Asian technology ecosystems and policymakers, China's regulatory approach offers instructive precedent. The region hosts substantial populations of young users consuming artificial intelligence services, and many Southeast Asian governments remain in early stages of formulating AI governance frameworks. China's experience suggests that regulatory attention concentrates particularly on services designed to simulate emotional intimacy, where psychological risks concentrate most acutely. The combination of digital restrictions and hardware-focused ethical standards indicates that Chinese policymakers view artificial intimacy as a governance priority spanning multiple technological domains.
The retreat by Chinese platforms arrives as these services have achieved remarkable market penetration. Doubao in particular has cultivated user bases comparing favorably with established messaging and social applications, suggesting that regulatory actions affect genuinely consequential products with substantial user dependency. The shift toward standalone companion applications—rather than outright prohibition—indicates regulatory pragmatism: authorities acknowledge market demand for these services while attempting to implement structural safeguards limiting psychological harm.
The regulatory framework also reflects evolving attitudes toward data governance within Chinese technology policy. By restricting conversation data usage for model training, Beijing is establishing precedent that intimate user disclosures constitute a protected category distinct from conventional user-generated content. This distinction carries implications for how technology companies globally might consider the governance of sensitive personal information, particularly data arising from users' interactions with emotionally-engaged AI systems.
Longer term, these regulatory actions may influence how artificial intelligence companies globally design systems incorporating emotional engagement capabilities. If Chinese regulatory restrictions on AI companion services prove effective in reducing psychological harms without substantially dampening innovation, comparable jurisdictions may adopt parallel approaches. Conversely, if restrictions generate compliance costs that disadvantage Chinese firms internationally, regulatory pressures elsewhere might moderate.
