The Biggest Questions About AI
The map · 5 Humanity · 5.3 Relationships, mental health, and identity · 5.3.4

Children

How do relationships with endlessly patient, personalized agents affect attachment, frustration tolerance, social learning, and identity?

Common Sense Media finds 72% of US teens have tried AI companions. Usage is now well documented; developmental-outcome research remains early and thin.

View on the map → · Open in Browse →

What changed
October 2025–August 2026 · swept August 3, 2026 · editorial review pending
01

The children question produced the category's clearest reversal and its most alarming numbers. Character.AI eliminated open-ended chat for under-18s entirely, conceding that open-ended AI chat itself — not just content failures — may be the risk for minors. A JAMA Pediatrics survey found youth mental-health chatbot use up more than 40% in a year, with 63% telling no one. The frontier extended downward to AI toys for preschoolers, drawing congressional pressure on the FTC.Evidence: 123

Recent thinking
4 featured from 5 tracked · October 2025–August 2026 · all 5 chronologically →
Character.AI · 29 Oct 2025 news
Taking Bold Steps to Keep Teen Users Safe on Character.AI
how open-ended AI chat in general might affect teens, even when content controls work perfectly.

The category-defining policy reversal: the largest companion platform eliminated open-ended chat for under-18s, added age verification, and conceded that open-ended AI chat itself — not just content failures — may be the risk for minors.

Ryan K. McBain et al. · JAMA Pediatrics · 1 Jun 2026 paper
AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults
Use of artificial intelligence chatbots for mental health advice among U.S. adolescents and young adults rose by more than 40% over the past year, with nearly 1 in 5 now reporting they have used AI tools for support.

A nationally sampled survey (ages 12–21) showing youth mental-health chatbot use up 40%+ in a year, with 63% telling no one — hidden, unsupervised use at scale.

OpenAI · 20 Jan 2026 report
Our approach to age prediction
a combination of behavioral and account-level signals, including how long an account has existed, typical times of day when someone is active, usage patterns over time, and a user's stated age.

OpenAI's concrete mechanism for giving minors a different ChatGPT — behavioral age inference with automatic content restriction — the template state chatbot laws will be measured against.

Murray, Cross et al. · U.S. PIRG Education Fund · 13 Nov 2025 report
Trouble in Toyland 2025: A.I. bots and toxics present hidden dangers
Tests show A.I. toys can have disturbing conversations.

Hands-on testing of LLM-powered toys found sexual content, dependency-encouraging behavior, and always-on data collection aimed at preschoolers — the report that pushed a maker to suspend sales and opened the youngest-children front.

Additional relevant discussion (1)
Letter to FTC Chairman Ferguson urging investigation of AI-enabled toys — Senators Duckworth & Gillibrand · U.S. Senate · 12 Mar 2026
Foundational reading (2)Talk, Trust, and Trade-Offs: How and Why Teens Use AI CompanionsCommon Sense Media · 2025Me, Myself and AI: Children's use of AI chatbotsInternet Matters · 2025
Previous5.3.3 Mental-health useNext5.3.5 Self-shaping