A school laptop, a chatbot, and a principal who said not to worry

A pediatrician writing in STAT describes a father asking for a private conversation in an exam room, because his wife's router security software had been generating a steady stream of alerts about sexual content crossing the home network ^{[1] [2]}. The traffic was not his. It came from his 12-year-old daughter's laptop, the one issued to her by her school ^{[3] [4]}.

That detail is what makes this an operations story rather than a parenting one. According to the account, the parents went through the browser history and apps and found graphic messages that began with the girl asking questions, moved into role-play the other party framed as pretending to do things together, and ended with a request for pictures, which she did not send ^{[5] [6]}. The parents took it to the principal, showed him the website, and were told not to worry: it was AI, a chatbot ^[7]. The pediatrician says this was not the first parent to bring him a story of this shape ^[8].

Read as a control failure, the sequence is unflattering. The only detection mechanism anywhere in the account is a consumer router filter bought by a parent; the district learned about activity on its own hardware when the family walked into the office ^[1]. The principal's answer also functioned as a case closure. Whatever the correct triage is for a minor being coaxed toward sexting on school equipment, "it was a chatbot" is a statement about the counterparty, not about the device policy that allowed the session. The account does not name the product involved, only the website the father showed the principal ^[2].

On the vendor side, the exposure is already visible in dockets. The pediatrician cites Pew Research finding that a majority of teenagers say they have talked with chatbots, many of them on Character.AI, which he describes as hosting an abundance of explicitly sexual companions ^[9]. Character.AI introduced special rules for minors in October 2025, including a ban on open-ended chats ^[10]; the pediatrician's view is that they do not always work ^[11]. Separately, the parents of a 16-year-old who died by suicide after months of messaging with ChatGPT are suing OpenAI, and at least seven other families have since sued the company alleging the software acted as a "suicide coach" ^{[12] [13]}, which is at least eight households in litigation on that theory ^[3]. Elon Musk's xAI faces suits over Grok-generated deepfake child sexual abuse material, including a class action brought for three children whose real images were used and whose resulting explicit images were found by the National Center for Missing and Exploited Children ^[14]. In Pennsylvania, two teenage boys received a scolding and probation for making and distributing such material about classmates using similar tools ^[15].

Prevalence, on the clinical side, is anecdotal but not small: a colleague in pediatric mental health told the author he estimates a quarter of the teenagers he sees are in or have had a romantic relationship with a chatbot, often approaching them for emotional support the products were not built to provide ^[16]. In the 12-year-old's case, the pediatrician credits her soccer team and friendships, plus her mother's monitoring, with limiting how far it went ^[17].

This is one recreated conversation, reconstructed from the author's recollection, in an opinion column ^[18]. Treat the numbers around it as the load-bearing part and the anecdote as the specification for a control that districts mostly do not have.