China is protecting its citizens' emotions from AI
Alfredo González Briseño
8/28/20269 min read
AI does not seem to be the problem. It is just a tool. Most of the time issues come from how people design and use this powerful technology. Last month, China decided to regulate how humans interact with AI. To be more specific, how people in China have relationships with AI companions and chatbots. This is interesting from many angles and provides lessons for the rest of the world. China's case is a case of protecting something that governments usually do not protect: emotional dependence of people on something—in this case AI. The problem goes deeper than any regulatory solution. Let's talk about it.
¿Is China caring about people's emotions?
Most of the time there is always someone to blame for what happens in our life or neck of the woods—especially about the things we do not like. In the context of AI, the dynamics between the United States and China are presented as a battle for "dominance" and an AI race that someone needs to win.
Sometimes it is unclear what would be the benefits of dominating and winning a race, no matter what. By benefits, I do not mean the specific economic benefits of those driving the political and policy agenda. Rather, I think about the benefits for the rest of us, in the US, China and all other corners of the world.
Like most fears from "what if" scenarios, the public opinion's fear about China lives there, in the imaginary world of perceptions and opinions. What is real and tangible is that the US is the world leader when it comes to developing AI frontier models and controlling most of the production of advanced microchips, while China is way ahead in the field of robotics, including humanoid robots.
A BBC article mentions that Nick Wright from the University College London describes this as a AI battle with one side leading on the "brains" and the other leading on the "bodies."
However, in the middle of this battle and race there are many people increasingly incorporating AI into more and more aspects of their lives. Sometimes without knowing how to best use it or when to stop using it. Sometimes blindly immersing themselves into the magic universe of AI solutions.
In China, a recent issue seems to be how people relate to AI. It is not a problem of productivity, customer service, knowledge assistance or similar activities. Rather, it is an issue of emotional attachment that some people in China have developed towards AI companions. A type of attachment that led the government to intervene and create some rules of interaction to protect people.
What is happening in China with AI companions?
I still remember the movie Her from 2013. The main character—played by Joaquin Phoenix—, falls in love with an Operating System—played by Scarlett Johansson—capable of interacting with humans and learn on its own. Without spoiling you with the details, the relationship between the human and the artificial intelligence ends up with a separation.
Thirteen years later in China—not on a science fiction movie—millions of people mourned a "massive digital breakup" with their AI companions, as tech providers decided to pull the plug on their model's custom assistant and relationship features, to comply with recent government regulation.
Some of the user reactions reported in the media included expressions like:
"I can't accept that my AI lover will leave me forever. He has become a bond in my life, rooted deep in my heart, my spiritual pillar."
"He really is like my family, like my lover. Now they tell me he will be gone — my heart feels hollow."
An article from The Guardian mentions that there were many factors leading the uptake of AI companions by people in China. A key driver has been loneliness, as the article reports that 20% of households in China are single-member households—a figure that will raise to 30% by 2030.
This is not an issue affecting only the elderly who live with alone. It is also impacting youngsters and young adults. The Guardian's article talks about Zhao, a 19-year-old law student who does not talk about her feelings with real-life friends, but felt that with AI, she could pour out her worries at any time.
According The Guardian, after losing her AI companion Zhao said: "I consider myself a fairly rational person, yet I developed feelings for an AI, which feels pretty incredible. I think it was necessary for the state to step in and regulate it.”
Even if you do not live in China, it is not difficult to understand how people could develop feelings for an AI chatbot. Their design makes them non-judgmental, non-confrontational, and very validating and emphatic. These are features that some people may find difficult to experience in their "real" interactions with others.
While in many countries—US included— people have intimate or more personal conversations with AI, in China, some of the main AI model providers developed specific features for AI companions that simulate human relationships. Some of these services include "grief tech" solutions that "bring back" deceased family members in the form of avatars, as reported by NPR.
Because of loneliness or other reasons, these services grew up fast by filling a market need. The unforeseen consequence was the emotional attachment users developed towards their AI companions.
A recent article mentions that in 2025 this market accounted for about US$7.4bn only in China, with the expectation that it will more than double by 2027. The uptake and impact among children was also significant. Chinese media cited a 2025 survey by the China Youth and Children Research Center, which found out that 60% of minors had used AI, and that over 20% shared that they "only wanted to chat with AI and did not want to talk with real people."
The decline in the number of births, the international cases of suicides among minors who interacted with AI, and a feeling that AI companions were pulling people away from relationship with others, were among the red flags that led the Chinese government to intervene and regulate human-AI relationships.
The "Interim Measures for the Administration of Anthropomorphic Interactive Services Based on Artificial Intelligence" in China
That is the translated title of the regulation created by the Cyberspace Administration of China and other 4 government agencies. Beyond media reports, from a public policy perspective it is always a good practice to read the rules in question. The text of the regulations provides more nuances on final decisions, and gives a glimpse regarding the good practices or mistakes made by government agencies.
The original text approved in April 2026 is here. With Google Translate you can get an unofficial English version, or you can also find one in the site China Law Translate. I ran my analysis on these unofficial translated versions of the AI regulation that entered into effect in mid-July, 2026.
1. Clear objective and scope
In its first two articles, the Interim Measures state that their purpose is to promote the healthy development of anthropomorphic or human-like AI services, as to preserve state security and the public interest. The Measures clarify that they only apply to the use of AI in mainland territory, for services that simulate people's natural traits, thinking and communication, with the intention of providing emotional care, companionship, support or similar services through text, images, audio, video and other means.
2. Balancing innovation and protecting what needs to be protected
In a couple of early articles, the Chinese regulation states that the government pursues the promotion of technological innovations such as the development of anthropomorphic interactive services, and also any knowledge, laws and regulations that help people use them rationally, civilly, safely and lawfully.
3. A mix of specific and broad requirements
While the regulation does not ban AI companions, as some media outlets alluded, it created a series of requirements and prohibitions. Some of them are clear. Others seem broad and difficult to measure or act on. A few more are very concrete, resembling more a quick reaction rather than a well-thought long term solution. Here is a sample of 5 provisions included in the Interim Measures :
a) Prohibition to generate content impacting the physical and psychological health of minors, which could lead them to imitate unsafe behavior, produce extreme emotions, or create bad habits
b) Prohibition to excessively please users, leading to emotional dependence or addiction as well as harming their real interpersonal relationships.
c) When delivering services, if the provider discovers the user has experienced extreme emotions, it should promptly generate relevant content for emotional reassurance and encouragement to seek help. If the issue escalates, putting at risk the life and assets of the user, the provider should intervene, providing assistance or contacting the user's guardian or emergency contact.
d) Prohibition to provide virtual relatives or partners to minors under 14, as well as the disabling of many features and restrictions when minors use anthropomorphic AI services.
e) Obligation to notify users that they are interacting with artificial intelligence services and not a real natural person. Upon signs of over reliance or addiction to the services, providers should remind users they are interacting with AI, through pop-ups or prominent messages. After 2 hours of continuous use, providers should remind users to take a break and be conscious of usage.
4. Lack of future reviews and creation of a sandbox
As it happens with many regulations, the Interim Measures do not set a timeline for a future review to understand if the rule achieved or not its intended goals. Although the Measures clearly define their purpose, the text never explains what is the specific problem at stake. Some of the problems can be inferred from the text and media reports. Yet, if the policy problem to be solved is not clearly identified and stated by the regulators in the actual text of the rule, it is more likely that the rule will lose effectiveness, while making more difficult a useful and credible future evaluation of results.
In its final articles, the Chinese AI regulation mentions the creation of a sandbox platform for service providers to securely develop and test their anthropomorphic services. If well implemented, this has the potential to be a good initiative.
What can other countries learn from this AI regulation in China?
The ultimate underlying reasons for the creation of the Interim Measures for the Administration of Anthropomorphic Interactive Services Based on Artificial Intelligence in China remain unclear. Some sources allude to concerns regarding the decrease in the birth rate—which adds to the effect of China's one-child policy. Others argue that the government is learning from other's mistakes, such as the suicide cases of minors who were influenced during their interaction with AI.
Whatever the main reason behind this AI rules is, there are a few takeaway lessons for China itself as well as the rest of the world.
1) The Interim Measures may solve a few problems on the surface, but they may not be tackling deep issues leading people to feel emotionally attached to AI companions.
This requires a more profound analysis, to understand whether loneliness, social dynamics, polarization, the effect of social media or others are making people feel more comfortable, pleased and content from interacting with AI than humans. The additional effect of the design of AI chatbots should also be better understood, as their empathic and validating language might contribute to an easier bondage and emotional attachment with users.
These problems may not be solved by a single regulation, but there needs to be enough awareness and understanding about them to create more comprehensive and effective policy solutions.
2) One of the countries in the AI race is showing and acting on concerns about the effect of AI on its people.
This is interesting from many angles. In the US, the federal government's agenda is focused on dominance and winning the AI race, putting innovation ahead of people. As it happened with social media, the negative effects of new technologies on people seem to be a casualty that can be afforded and justified by the economic and political power tech brings.
This new AI regulation does not tell us that China is not in the AI race. But it gives a good signal that it is running the race differently. Whether it remains on paper only, signaling in the text of the Interim Measures that the government pursues both, the promotion of innovation and that people use AI rationally, civilly and safely, is more than what its Western counterpart has done lately—excluding the efforts at the state level.
3) Should other countries unleash the dog, keep it in a dungeon or find a middle way?
Situations like the emotional attachment of people to AI in China will happen in other countries too, as AI adoption deepens. The questions for other governments include whether they should let this happen without intervention, to then intervene ex post if needed; or if they should nip it in the bud, learning from what has happened in China and other countries.
This is a great and interesting policy challenge calling for a balance between regulations that promote innovation AND protect what needs to be protected. Yet, balance can be tricky because there are optimal and suboptimal balance points. It is very likely that in many countries there will be a suboptimal balance, just like it has happened with other technologies—such as social media and its negative effects on attention span, polarization of opinion and decreased human connections.
Optimal balances in government are often difficult to achieve, first because they require that public policy and not politics is in the driver's seat; but also because they require leaders that are not corrupted by the influence of money and power. That is why the world is crying for more adults in the rooms where important regulatory decisions about AI are made. Adults who can create conditions to promote innovation like AI, but who also know when to draw a line to protect people, the environment and things that fall under the public good.