AI Companion Apps Are Scaling Fast — and Pushing Emotional AI Ethics Onto the Table

Character.ai, Replika and domestic products such as Xingye and Zhumengdao have turned long-running chats with a model into their own category. User bases have clearly expanded over the past two years, which has also pushed emotional dependence, minor protection and identity disclosure onto the regulatory table. This site previously covered Character.AI's ban on free-form chats for under-18s — a concrete event on that line.

The Fight Is Over Design, Not Chit-Chat#

The argument is not whether the bots can talk. It is what kind of relationship the product is designed to be: whether it encourages treating a model as primary emotional support, whether it offers romance simulation to minors, and whether conversations are reused for training. Psychologists and teacher groups keep warning that substituting real social contact can weaken offline relationships, especially for adolescents.

Platforms Tightening Their Own Rules#

In an official safety update dated September 3, 2026, Character.AI recalled that it had removed open-ended chat with Characters for users under 18, calling it one of the most significant safety decisions it has ever made (Character.AI blog).

On the regulatory side, several countries now require usage-time reminders, ban romance simulation for minors, and insist the product disclose that the user is talking to AI. Industry guidelines are still chasing the product cycle. For users, treating a companion app as a practice ground is one thing; treating it as the only emotional outlet is another.