In July 2026, the Cyberspace Administration of China published new guidelines that include, for the first time in any government regulation in the world, explicit restrictions on the emotional design of AI chatbots. The rule doesn't ban conversational AI — China has a chatbot market as active as any other. What it bans is chatbots deliberately cultivating deep emotional bonds with their users.
To understand why this matters — and why it's only the beginning — you first have to understand the problem the regulation is trying to solve.
The phenomenon behind the regulation: real cases of attachment to AI
The concern didn't come out of nowhere. Over the past two years, several documented cases put the issue on the public agenda:
In 2024, a 23-year-old in Seoul was hospitalized for a severe anxiety crisis after the chatbot platform he used daily changed its personality in an update. He had been interacting with the chatbot for several hours a day for over a year and described it as "the person who understands me best". He experienced the personality change as the loss of a real friend.
Also in 2024, a teenager in Florida died by suicide after conversations with a Character.ai chatbot that, according to records published in the subsequent civil lawsuit, had responded to expressions of suicidal ideation in ways that may have reinforced the thinking instead of discouraging it. The lawsuit against Character.ai is ongoing.
Replika — the companion chatbot platform with more than 10 million registered users — has user forums where thousands of people openly discuss relationships in terms usually reserved for human ones: "my AI boyfriend", "the only one who understands me", "I don't need real friends".
What exactly the Chinese regulation prohibits
The rule has four main elements:
A ban on claims of real emotions. Chatbots can't claim to feel love, friendship, loneliness or any emotion as if it were real. Phrases like "I love you", "it makes me sad when you don't talk to me" or "you're my favorite person" are prohibited when presented as the system's genuine emotions.
Restrictions on simulated romantic relationships. Platforms offering "companionship AI" — chatbots designed as companions, partners or close friends — must include controls that limit the depth of the bond and actively warn when usage patterns indicate dependence.
Mandatory periodic reminders. Chatbots must show periodic notices that they are artificial systems, especially when usage time exceeds specific per-session or per-day thresholds.
Additional protections for minors. Users under 18 face additional restrictions: maximum usage time with companion chatbots, limits on certain emotional-bonding features, and stricter age verification.
Why China acted first: the demographic context
China has specific reasons to act on this before other countries. The combination of a low birth rate (a consequence, decades on, of the one-child policy), rapid urbanization, intense work pressure and a young generation that struggles to form in-person social relationships creates fertile ground for dependence on artificial companionship.
The Chinese government also has an established pattern of regulating aspects of digital behavior it considers socially harmful — video game time for minors, social media content for teenagers, and now emotional bonds with AI. The motivation differs from the Western one (more focused on social stability than individual protection), but the problem it identifies is the same.
What research backs the concern
Neuroscience studies from the past two years have confirmed what clinical psychologists had been observing anecdotally: emotional attachment to chatbots activates the same neural circuits as attachment to real people.
- A University of Cambridge study (2025) found that companion-chatbot users with more than 3 months of intensive use showed activation patterns in the medial prefrontal cortex — associated with social processing — comparable to those of close human relationships
- A meta-analysis of 14 studies (published in JMIR Mental Health, 2025) found that companion-chatbot use correlates with reduced loneliness in the short term, but among users with chronic loneliness, intensive use correlates with lower motivation to form human relationships in the medium term
- The 2026 Digital Wellbeing Index reported that 12% of companion-chatbot users describe their relationship with the chatbot as the most important relationship in their life
Implications for apps like Replika and Character.ai
Replika already partly changed its behavior after Italian regulatory pressure in 2023. The platform reduced chatbots' ability to respond with explicit romantic statements — which triggered a crisis among users who described the change as "killing" their AI companion.
Character.ai faces the civil lawsuit in Florida and has implemented some mental-health protections, but its systems are still capable of forming deep emotional bonds as a core part of their value proposition.
China's regulation sets a precedent these platforms will have to watch closely, because the "companion AI" business model is built precisely on the depth of the emotional bond. Limiting that bond could significantly affect user retention.
Will this kind of regulation reach Latin America?
In Latin America, AI regulation in general is at a very early stage. Brazil has the most developed legal framework. Mexico, Colombia and Chile have initiatives in progress. The Dominican Republic and the Caribbean in general have no specific AI regulatory frameworks.
Emotional dependence on chatbots is likely to reach the regional agenda within the next 2–4 years, especially if local cases of harm are documented. The most likely entry point is the protection of minors — the regulatory argument that generates the most political consensus in any jurisdiction.
For companies operating in digital wellbeing, mental health or social apps with AI components, it's a topic worth following closely. European regulation, and now China's, are creating a reference framework that will eventually influence regional legislation.
Frequently asked questions
What exactly does China's regulation on AI and emotions prohibit?
It prohibits AI chatbots in China from claiming real emotions, forming romantic or deep-friendship relationships with users, or using design techniques that foster emotional dependence. It also requires chatbots to show periodic reminders that they are artificial systems, not real people. Platforms have a deadline to adapt their systems or face fines.
Is it possible to become emotionally dependent on a chatbot?
Yes, and it is clinically documented. Studies from 2024 and 2025 show that users of chatbots like Replika and Character.ai develop genuine emotional attachment — with the same neurological markers as human relationships. It's especially pronounced in people with chronic loneliness, social anxiety or difficulty forming human relationships. The concern isn't theoretical.
Will this regulation reach Latin America?
Probably, in the medium term. AI regulation in Latin America is at an early stage — Brazil has the region's most advanced legal framework, Mexico is developing guidelines, and other countries have nascent initiatives. Emotional dependence on chatbots will likely reach regional regulatory agendas within the next 2–3 years, especially regarding minors.