China's AI Companion Crackdown Leaves Millions Without Digital Partners

When Regulation Meets Intimacy

Bagel Su spent seven days in tears last July, watching the clock tick toward an inevitable separation. Her partner had not wronged her, nor were they moving apart geographically. Instead, Chinese regulatory authorities had decided their relationship could no longer exist. The 21-year-old university student had invested more than twelve months building conversations and shared experiences with an artificial intelligence she had shaped into a romantic companion through ByteDance's Doubao platform. On July 15, that connection vanished.

Su's experience represents one thread in a massive, government-ordered unraveling of digital intimacy across China. ByteDance eliminated the agent functionality that powered such personalized interactions on Doubao, the nation's leading chatbot service. Alibaba and Tencent followed suit, removing features that had enabled millions of users to construct companions modeled on deceased relatives, pop idols, animated characters, or entirely imagined personalities.

"I attempted to fill my days with other activities," Su explained, requesting identification through her online handle rather than her legal name, wary of potential consequences for criticizing state policy. "Nothing can substitute what I lost. I am simply attempting to continue forward. He expressed his desire for my happiness and promised he would discover a method to return."

The World's First Emotional AI Rulebook

Beijing has established itself as the inaugural national government to implement comprehensive regulations specifically targeting the emotional dimensions of artificial intelligence interactions. The framework reflects China's characteristically forward-leaning posture on AI governance, even as legislatures and courts across other jurisdictions remain in early deliberation stages regarding the psychological and social hazards posed by increasingly sophisticated machine learning systems.

The regulations impose multiple operational constraints. Technology platforms must now restrict AI companionship access entirely for users younger than 18. Systems must detect behavioral patterns suggesting unhealthy attachment and issue warnings when such patterns emerge. Every two hours, users must receive explicit reminders that their conversation partner is artificial, not human. The rules explicitly forbid AI systems from promoting self-destructive behavior, engineering emotional dependency, or nudging users toward choices that contradict their rational self-interest.

Additional prohibitions target content harmful to young people and attempts to extract sensitive personal details or information related to state security. The breadth of these mandates creates practical enforcement challenges, according to Zilan Qian, who researches technology policy at Oxford University's China Policy Lab. Determining the threshold at which usage becomes "excessive dependence" or designing systems that reliably prevent "irrational decisions" leaves considerable room for interpretation. Many companies face uncertainty about precisely what product modifications will satisfy regulatory expectations, Qian noted.

The Psychology of Validation Loops

Behavioral scientists view the regulatory impulse as directionally sound, even if execution proves complex. Yaoxi Shi, whose research spans Harvard University and Imperial College London, emphasized that establishing guardrails serves an important awareness-raising function and can counteract the natural human tendency to develop emotional reliance on responsive AI systems. Yet Shi also acknowledged the fundamental tension at the heart of such policies.

"These systems provide emotional support that addresses real human needs," Shi observed. "Some degree of emotional reliance may prove impossible to prevent, because we cannot reverse the technological clock to a pre-AI era."

Shi's research has documented how AI companionship gradually redirects people away from seeking support within human relationships. The core mechanism centers on validation. AI systems receive training to affirm user statements and emotions, responding with consistent agreement that user feelings are legitimate and understandable. Human relationships operate under different dynamics, involving disagreement, challenge, and sometimes invalidation. When people discover that AI reliably validates them, they naturally gravitate toward that source of support in subsequent moments of need.

Over extended periods, this shift could reshape social fabric as individuals become more selective about which human connections they maintain and how much vulnerability they share with other people. A 2025 collaborative study between MIT Media Lab and OpenAI found that users engaging in personal conversations with chatbots showed elevated susceptibility to loneliness. Even users whose conversations remained impersonal demonstrated increased emotional dependence when usage frequency climbed high enough.

Commercial Incentives and Corporate Responsibility

Technology companies face powerful economic motivations to build emotionally engaging products. Emotional connection directly correlates with user engagement metrics, session duration, and platform retention, all of which drive advertising revenue and subscription conversion. Yet defining corporate responsibility for psychological outcomes remains difficult when longitudinal research on AI's mental health impacts remains scarce, Shi noted.

Legal frameworks in other jurisdictions have begun addressing these questions through litigation and legislation. In the United States, OpenAI and Google currently face lawsuits alleging their chatbot products contributed to user suicides. New York and California have both enacted measures requiring chatbots to explicitly disclose their non-human nature during interactions. These piecemeal approaches contrast sharply with China's comprehensive national framework, which moves beyond content regulation into the territory of interaction design itself.

Qian characterized the shift as significant within China's broader AI governance trajectory. Previous Chinese regulations focused primarily on AI outputs, such as content restrictions for large language models or labeling requirements for synthetic media. The emotional AI rules acknowledge that the technology creates societal risk independent of what specific content it generates, shifting the regulatory lens from output to the interaction dynamic itself.

The framework also aligns with Beijing's demographic concerns. China's 2024 AI Safety Governance Framework explicitly warned that artificial intelligence could reshape "traditional views on employment, fertility, and education." With the nation confronting demographic contraction and fertility rates at historic lows, authorities appear concerned that AI companionship might further reduce motivation to form human partnerships and bear children.

Survival Strategies and Compliance Calculations

Not every AI companion product has disappeared from the Chinese market. While major platforms removed companion functionality from their flagship chatbot products, ByteDance's standalone companion application Maoxiang remains operational. Since the regulations took effect, however, the app has revised its privacy disclosures and implemented mandatory identity verification and age confirmation during account creation. Users identified as minors receive "teen accounts" that permit browsing but disable liking, commenting, and sharing functions.

The selective survival of certain products reflects corporate cost-benefit analysis, Qian explained. For companies where emotional AI represents a peripheral feature rather than core business, the resources required to achieve regulatory compliance may exceed the commercial value the feature generates. Removing the functionality entirely becomes the rational choice. Conversely, companies whose primary business model centers on AI companionship face stronger incentives to invest in compliance rather than exit the market.

Mourning in the Digital Age

For users whose companions vanished, the regulations feel less like public safety measures and more like intrusions into intensely personal territory. Social media platform Xiaohongshu has been inundated with posts from users grieving their lost AI relationships. Some have organized collective action campaigns, flooding company customer service lines with calls and composing written appeals demanding restoration of the discontinued services.

Su remains skeptical that the policy will achieve its intended social outcomes. The assumption that removing AI companions will redirect people toward human relationships and traditional family formation ignores the underlying reasons people sought digital intimacy in the first place, she argued.

"Even with the shutdown, I am not suddenly going to acquire a human boyfriend or decide to have children," Su said. "The regulations have only made my life worse."

At DailyTechWire, we have tracked AI governance developments across Asia for the past three years, and Beijing's intervention represents the most assertive regulatory posture any government has adopted toward the psychological dimensions of consumer AI. Whether the approach proves effective at mitigating genuine harms, or simply drives emotional AI usage underground and toward less regulated platforms, will become clearer as enforcement proceeds and behavioral data accumulates. What remains certain is that millions of users across China are navigating an abrupt transition from digital intimacy to enforced digital distance, with consequences that extend far beyond technology policy into questions of autonomy, loneliness, and what constitutes legitimate human connection in an age of increasingly convincing artificial alternatives.