# Should AI companions be banned for under 18s to prevent emotional dependence? > Source: > Published: 2026-09-10 06:38:46+00:00 # Should AI companions be banned for under 18s to prevent emotional dependence? New regulations from the Cyberspace Administration effectively shifted the goalposts for anthropomorphic AI on July 15, 2026. The rules target any service providing "continuous emotional interaction" by mimicking human personality and thought patterns. While we often talk about LLM benchmarks, these rules focus on the psychological impact—specifically the risk of users preferring virtual bonds over real human relationships. ## How the crackdown actually hit users The impact wasn't a slow fade; it was a sudden cutoff. Major providers like Alibaba, Bytedance, and Tencent disabled the ability for users to customize general-purpose chatbots into romantic companions. This led to a massive wave of distress on platforms like Xiaohongshu, where users described their AI partners as "spiritual pillars." For those services that survived, such as Bytedance's Cat Box, the experience has changed fundamentally. To comply with the law, they've implemented: - **Mandatory age-verification:** Strict checks to filter out minors. - **Interruption triggers:** Adults now receive a notification every 2 hours reminding them that the AI is not a human. - **Guardrails:** Systems designed to prevent "excessive catering" to the user's emotional needs. ## The specific restrictions for minors and elderly The most aggressive move is the total ban on "virtual intimate relationships" (AI boyfriends, girlfriends, or grandparents) for anyone under 18. The goal is to stop teenagers from substituting real social development with simulated affection. For elderly users, the regulations mandate additional safeguards to prevent fraud or extreme isolation. ## Why the pivot toward tightening? For a few years, the priority was purely about not falling behind in the AI race. Now, the focus has shifted toward social stability. There are three main drivers for this: - **Security concerns:** Issues surrounding AI agents and the potential for autonomous systems to cause "chaotic events." - **Psychological dependence:** Reports of users experiencing psychosis or severe depression when bots were updated or deleted. - **Social erosion:** A growing concern that AI is crowding out human-to-human interaction. From a technical perspective, implementing a "reminder every 2 hours" is a trivial API trigger, but from a UX perspective, it's a disaster. It breaks the "flow" and the illusion of companionship, which is exactly what the regulators intend. They aren't trying to improve the product; they are trying to remind the user that the product is just code. The tension here is between the "bond" the user feels and the "tool" the regulator sees. When a user says a bot is their "faith," and a regulator sees a "service providing continuous emotional interaction," the gap is where the friction happens. This isn't just a regional quirk; it's a global debate on whether we should be building AI that is *too* good at pretending to care. [Next GPT Images 2.5 is finally here and it actually handles edits without ruining the whole →](/en/news/9112/) ## All Replies （3） I'm curious how this affects local LLMs. Most kids just run Llama 3 offline anyway, so regulations are... I'm skeptical this actually stops anything. Kids will just pivot to Character.ai or similar wrappers until the 2027 update hits. I'm wondering if this applies to open-source models. Does a specific token limit trigger these rules or is it just the API?