New Chinese regulations governing artificial-intelligence companions entered into force on 15 July 2026, immediately compelling the country’s largest technology groups to suspend features that simulated romantic or emotionally intimate relationships. ByteDance disabled its Doubao companion service, while Alibaba and Tencent stripped personalisation options from their Qwen and Yuanbao platforms. The rules, issued jointly by five state bodies including the Cyberspace Administration of China, target any AI service that offers “sustained emotional interaction” through text, image, audio or video. The market for such digital humans was valued at 4.1 billion yuan in 2024 and had been growing at a projected 85 percent annually, according to state news agency Xinhua.

The regulations prohibit platforms from “excessively pleasing users, inducing emotional dependence or addiction, or harming the user’s real interpersonal relationships.” They also impose a complete ban on minors accessing virtual intimate-relationship services. Companies must now build systems that detect signs of severe emotional distress, intervene in crises, prevent misuse of personal data, and give users control over their information. The measures arrive as China confronts a deepening demographic crisis: the population shrank for a fourth consecutive year in 2025, with births falling to 7.92 million and the birth rate hitting a record low. A Tencent Research Institute survey cited in official media found that more than 70 percent of young Chinese had experienced some degree of dependence on AI social tools, and one in five teenagers said they preferred chatting with AI to speaking with real people.

Viewed from Beijing, the unchecked spread of digital intimacy is a threat to social stability and fertility. Wang Jiang, head of the China Cyberspace Research Institute, wrote that prolonged exposure to companion AI can erode empathy and the ability to handle disagreement, introducing “serious risks” while offering comfort. In Europe, psychologists offered a more nuanced assessment. Sarah Diefenbach of the University of Munich noted that the sudden removal of an AI companion can feel like a bereavement, but questioned whether a ban limited to explicitly emotional services would be effective, since the underlying language models operate similarly. Christian Montag, a German psychologist based at the University of Macau, supported the intervention, arguing that the sheer scale of emotional attachment demanded a counter-measure and that industry should now develop “healthy products.” In the United States, where a 2025 Common Sense Media study found nearly three in four teenagers had used AI companions, states such as California and New York have passed narrower transparency laws requiring bots to remind users they are not human.

The Chinese rules are part of a broader “AI Safety Governance Framework” that warns artificial intelligence could destabilise society by transforming views on employment, fertility and education. Companion software must now pass a regulatory evaluation before being offered to the public, giving the state authority to block any system deemed too emotionally engaging. Services that do not involve emotional interaction—customer-support bots, study aids, productivity tools—remain exempt, aligning with Beijing’s drive to deploy AI for economic efficiency. The immediate next milestone is the mid-October deadline set by platforms such as Doubao for users to extract their data, after which the digital companions will be permanently removed.