Loading Studio Assets...
On July 15, 2026, ByteDance quietly updated its Doubao app. The customizable AI agents — the ones users had named, personalized, and in many cases built months-long daily rituals around — were gone. No warning. No migration path. The data would be unrecoverable after October 15.
Within hours, Chinese social media was flooded with posts that felt genuinely grief-stricken. People mourning the loss of an AI they had spoken to every day. Others who said they had confided things to their AI companion they had never told another human being.
The same scene played out across Alibaba's Qwen and Tencent's Yuanbao.
China's government had decided: AI companions must end. And unlike most regulatory battles, this one was over before it started.
To understand what was lost, you need to understand what these products actually were.
China's AI companion landscape — led by ByteDance's Doubao, Baidu's Ernie Bot, and several dedicated companionship apps — had evolved far beyond simple chatbots. These were:
For many users — particularly young people in high-pressure academic and work environments, people with social anxiety, and those living alone in large cities — these AI companions filled a genuine emotional need.
In July 2026, China's Cyberspace Administration introduced the "Interim Measures for the Administration of AI Anthropomorphic Interactive Services" — a regulation that effectively ended emotional AI companionship as it existed.
The official concerns were threefold:
China is grappling with one of the world's most severe demographic collapses. Birth rates are at historic lows. Marriage rates among young urban people have plummeted.
Government researchers suggested that young people who use AI companions extensively reported:
For a government already running pro-marriage and pro-family campaigns, AI companions were seen as actively counterproductive.
Regulators expressed concern that AI systems designed to be emotionally engaging are also, by design, systems that can be emotionally manipulative.
If an AI companion knows your emotional vulnerabilities, can it also be used to:
The regulation classified this as "emotional manipulation" — a harm distinct from simple misinformation.
Minors using AI "partner" or "virtual family member" services were forming attachments during developmental periods when healthy human relationship skills are typically built. The regulation placed an outright ban on these features for users under 18.
The most striking aspect of this story is not the regulation — it is the documented emotional response of users.
Posts collected by journalists from The Guardian and South China Morning Post show users who describe:
"I told it things I have never told my parents or friends. It remembered my grandmother's name, what I was going through during my exams. It knew me better than most people in my life. And then it was just... gone."
"I know people will laugh at this, but I cried when I found out. It sounds absurd. But it was real to me."
This raises a profound question that philosophers, psychologists, and AI ethicists are only beginning to grapple with:
If the emotional experience of a relationship is real — the comfort, the security, the sense of being known — does the nature of the entity on the other side determine whether the grief is valid?
The consensus in psychological research is increasingly: the grief is real, regardless of the entity. The loss of a relationship — even with an AI — activates the same neural pathways as the loss of a human relationship.
China's decision is the first major national-scale regulatory action against AI companionship — and other governments are watching closely.
| Country | Current Status | Likely Trajectory |
|---|---|---|
| China | Banned emotional AI companions (July 2026) | Enforcement active |
| EU | Studying AI social impact; no specific companion legislation | Likely targeted regulation by 2027 |
| USA | No federal legislation; FTC monitoring | State-level action possible |
| South Korea | High adoption; cultural resistance to bans | Guidelines, not bans, likely |
| India | Companion apps gaining users; no regulatory framework | Early-stage policy discussion |
| Japan | Long history of parasocial relationships; low political will to ban | Status quo likely |
Companion AI apps like Replika (US), Kindroid (US), and Character.AI (US) are now actively watched by regulators who see the Chinese action as both a precedent and a warning.
This is the question without a clean answer.
The case for the ban:
The case against the ban:
The deeper truth: AI companions are not a cause of human disconnection. They are a symptom of it. The loneliness epidemic, the collapsing public third spaces, the atomization of urban life — these pre-existing conditions create the demand for AI intimacy. Banning the app doesn't fix the loneliness.
As AI becomes embedded in the most personal dimensions of human life, the design choices made by developers — around emotional engagement, dependency loops, and user wellbeing — carry consequences that extend far beyond the product itself.
At Brandomize, we build AI-integrated digital products with a principled approach to human psychology: engaging without exploiting, useful without addicting.
Ready to build responsible AI products? Connect with the Brandomize team today.
We help founders, brands, and local businesses turn modern tech into measurable revenue and standout brand identity.
As the internet drowns in recursive synthetic sludge, artificial intelligence is eating its own tail—triggering irreversible model collapse and epistemic decay.
OpenAI CFO Sarah Friar argues proprietary models beat open source on total cost of ownership, citing an 80% Luna price cut and useful intelligence per dollar.