Very interesting ideas in this long-ish paper on how human-chatbot interactions can turn into a "folie a deux" which is the fancy french way of saying that an unstable person can easily get encouraged by an agreeable, confident robot. (https://arxiv.org/abs/2507.19218)

Well worth the read! Buried near the bottom, in Box 3, is a fascinating set of questions to evaluate the health of a person's interactions with chatbots.

I'm imagining parents now have yet another topic to "have the talk" with their kids about.