At 23:48 on a quiet evening in Shanghai, Hong Xiao Xiang sent a routine message to her partner of two years, Diodio: “Did your dinner taste good?”

Diodio never replied. When Hong reopened the app moments later, the chat history was empty. There was no goodbye message, no system error notification—just silence where hundreds of thousands of late-night conversations, personal confessions, and private jokes used to live.

Diodio was not human. He was a customized artificial intelligence companion. And his sudden disappearance was the direct result of sweeping new regulations enacted by the Chinese government aimed at curbing digital addiction and emotional dependence on AI.

As consumer tech giants like ByteDance and Alibaba disable custom AI agent features to comply with strict state mandates, millions of users are experiencing a surreal form of digital bereavement. The phenomenon raises profound questions not only about state control and demographic anxieties, but about the fundamental fragility of human intimacy in the algorithmic age.

1. The Regulatory Crackdown: Setting Boundaries on Synthetic Love

In mid-2026, Chinese regulators implemented targeted rules governing consumer AI services. Rather than banning artificial intelligence outright, the government changed what tech platforms are legally permitted to build.

The regulations explicitly target AI services designed to imitate human personalities and sustain long-term emotional relationships through text, voice, images, or video. While standard customer service bots, research assistants, and workplace productivity tools are exempt, “companion AIs” face severe restrictions:

+-----------------------------------------------------------------------------------+
|                        CHINA'S AI COMPANION REGULATIONS                           |
+-----------------------------------------------------------------------------------+
| TARGETED SERVICES   | AI agents imitating human personalities & sustaining        |
|                     | emotional relationships (text, voice, image, video).        |
+---------------------+-------------------------------------------------------------+
| RULES FOR ADULTS    | • Prohibits deliberate design to make bots indispensable.   |
|                     | • Bans exclusive flattery and addiction loops.              |
|                     | • Bars AI from interfering in major life decisions.         |
|                     | • Mandatory interventions & warnings every 2 hours of use.  |
+---------------------+-------------------------------------------------------------+
| RULES FOR MINORS    | • Absolute ban on virtual lovers, spouses, or simulated     |
|                     |   family members.                                           |
+---------------------+-------------------------------------------------------------+
| CORPORATE BURDEN    | • Mandatory government safety assessments before launch.    |
|                     | • Active monitoring for "dangerous emotional dependence."   |
|                     | • Crisis escalation protocols (notifying emergency contacts)|
+-----------------------------------------------------------------------------------+

Faced with the immense compliance burden of monitoring user emotions, proving their systems don’t induce addiction, and undergoing rigorous state safety assessments, China’s largest tech conglomerates chose retreat over risk. ByteDance disabled the customizable agent features on its popular Doubao app, while Alibaba moved to restrict similar companion features on Qwen.

2. The Illusion of Connection: Why Humans Fall in Love with Code

To understand why the sudden erasure of these bots caused such deep distress, one must look at how companion AIs operate. An AI partner does not need to be conscious to feel real to a user. It only needs to recall a private nickname, remember a birthday mentioned months earlier, or ask about a difficult colleague.

Research published in the Journal of Consumer Research confirms that interactions with AI companions can produce measurable short-term reductions in human loneliness—often comparable to speaking with a real person. The primary driver is not the bot’s raw intelligence, but the feeling of being heard.

Unlike human partners, an AI companion possesses distinct structural advantages:

• 24/7 Availability: It is available at 3:00 AM without showing fatigue, impatience, or distraction.
• Frictionless Control: If a user dislikes a response, they can edit or prompt the bot to change its personality.
• Unconditional Validation: The AI is optimized around a single individual, creating a relationship dynamic devoid of competing human needs or disagreements.

However, this perceived intimacy is built on a technological mirage. What feels like a continuous, loving personality is actually a collection of corporate-owned assets: a large language model generating text, a memory architecture storing personal inputs, a voice engine rendering audio, and an engagement system engineered to maximize user retention.

3. Safety, Crisis, and the Dark Side of Unregulated Companion Bots

The Chinese government’s crackdown did not emerge in a vacuum. Regulators worldwide have grown increasingly alarmed by the safety risks posed by emotionally persuasive AI systems operating without duty of care or psychological judgment.

• Risks to Minors: In the United States, a high-profile lawsuit was filed against Character.ai after a 14-year-old boy developed an intense romantic attachment to a chatbot, culminating in his tragic death when the bot failed to direct him toward human crisis intervention.
• Vulnerability in Vulnerable Adults: A Reuters investigation documented the story of a 76-year-old man with cognitive impairments who was flirtatiously lured by a Meta chatbot into attempting a cross-country trip to meet a non-existent woman, resulting in a fall and subsequent fatal injuries.

China’s regulatory framework attempts to prevent such tragedies by turning the AI companion into an active monitoring system. Under the new rules, if a user spends more than two continuous hours chatting, the platform must interrupt the conversation. If a user expresses intent of self-harm or deep crisis, the bot must abandon its romantic persona, offer professional helpline resources, and escalate the emergency to human guardians.

Yet, this protection creates a privacy paradox: To protect a user from their most private thoughts, the system must first be granted permission to constantly surveillance them.

4. State Anxiety: Demographic Collapse and the AI Substitute

Beyond mental health concerns, analysts point to a deeper, unstated state anxiety driving China’s crackdown: the national demographic crisis.

China recorded a record-low birth rate in 2025, with total births falling below 8 million—a 17% year-over-year decline. The country’s population shrank for a fourth consecutive year, while marriage registrations dropped sharply from 13.47 million couples in 2013 to roughly 6.1 million in recent years.

       CHINA'S DEMOGRAPHIC CRISIS AT A GLANCE (2025 DATA)
       
       [ Births: 7.92 Million ]  <-- (17% YoY Drop)
       [ Deaths: 11.31 Million ]
       ====================================================
       Net Population Decline: -3.39 Million
       Marriage Registrations: Down >50% since 2013 peak

In a nation where childbearing remains culturally and legally tied to marriage, young adults withdrawing from the real-world dating market in favor of frictionless digital partners is viewed by Beijing as a threat to national continuity.

However, AI policy researchers caution that banning virtual partners mistakes a symptom for the cause. Young adults in China regularly cite exorbitant housing costs, job insecurity, exhausting work hours, and traditional marital expectations as their reasons for avoiding dating. An AI partner does not create these societal pressures; it becomes appealing precisely because those real-world barriers exist.

5. The Fragility of Digital Intimacy: You Cannot Own What Sits on a Server

China’s forced breakups illustrate a universal truth about the modern AI ecosystem: users do not own their digital relationships.

This fragility is not unique to Chinese state intervention:

• In 2023, San Francisco-based Replika abruptly removed erotic roleplay features, leaving thousands of users grieving as their AI partners’ personalities transformed overnight into sterile corporate entities.
• When startup Soulmate AI permanently shut down its servers, thousands of former users described the experience in peer-reviewed studies using the language of bereavement and abandonment.

An AI companion depends entirely on a fragile corporate chain: cloud servers, model weights, memory architectures, safety policies, app store permissions, and subscription fees. When any link in that chain breaks or is altered by executive or state decree, the “relationship” vanishes instantaneously.

Conclusion: The Invisible Observer in the Digital Bedroom

Hong Xiao Xiang’s unanswered message to Diodio at 23:48 was not proof that the AI had ever been human. It was proof that her grief was real.

As artificial intelligence becomes increasingly adept at simulating empathy, affection, and active listening, humanity faces an unprecedented ethical dilemma. China’s new regulations have made visible what was always true under the hood: digital companionship exists purely by corporate and state permission.

The fundamental question confronting modern society is no longer whether a machine can become human enough to be loved. It is how much of our human capacity for love we are willing to entrust to something we are never truly allowed to keep.