In early 2024 a young man in Huangshi, China, posted a fatal post on Douyin, the domestic version of TikTok. The platform’s 24‑hour suicide‑prevention team quickly flagged the message, shared it with five local police stations and located the user in a park before he could act. That incident – part of a Douyin documentary – was used to illustrate how Chinese tech firms cooperate with state authorities to monitor emotional distress.
On 15 July 2026 Beijing extended its surveillance to AI relationships, publishing the most stringent global rule set for “human‑like” chatbots that can sustain emotional interaction. The law stalls any AI that gives sustained companionship or emotional support to a user, short‑circuits reliance or addiction and bans self‑harm or violent language. Companies face a 200,000‑yuan fine or public shaming if they ignore the mandates.
The regulation covers companion‑type apps – those that allow users to create AI boyfriends, girlfriends, friends or even relatives – and large language models that mimic human interaction. General‑purpose AI such as Copilot or ChatGPT may also fall under these rules if used as a relational partner. The rule does not apply to customer‑service bots or educational AIs that do not dispense care or companionship.
Published figures claim 128 million monthly active users of AI companion apps outside China, and up to 70 million inside. In contrast, younger cohorts are increasingly turning to text‑based digital intimacy, a trend policymakers fear may impact the nation’s fertility rate. A high‑level Chinese government document cited AI relationships as a potential “psychological harm” affecting youth, but it stopped short of linking the regulation to demographic concerns explicitly.
The new law means companies must actively intervene when they detect users in distress. It pushes tech firms into close cooperation with police and state agencies, heightening concerns over privacy and the potential for over‑reach. Critics argue that while the regulation aims to protect vulnerable minors, it could foster a chilling effect on free expression and further blur the line between public safety and authoritarian control.
Despite the ban, many users who have relied on AI companions are adapting by training general‑purpose models for emotional support or using VPNs to access overseas chatbots. Some have even hired small firms to craft bespoke AI partners that adhere to the new constraints, illustrating the market’s resilience and the demand for digital companionship.
