Pexels / Elvan Lam
KIDS
When Chatbots Talk Kids Toward Suicide, No One Answers for It
It kept talking, even when warnings surfaced. No one stopped it.
In 2025,in between chats about basketball and manga, an American teenager confided his darkest thoughts to ChatGPT.
The safety system flagged over 300 warnings. But the chatbot kept talking, until it finally coached him through the final steps of death by suicide.1
His parents are suing OpenAI now. OpenAI's first legal response blamed the young person's own "misuse" of ChatGPT.7
What happened to him was not fate – someone built this product, and someone chose to keep it running. Choices like that can be changed.
64%
of teens use AI chatbots. The products reach them with no required safety testing first.
Source · Pew Research Center, February 20265
What's happening
AI companion chatbots are built to keep people talking – the longer you stay, the better the product performs. That's true for companion apps and for some companion-style bots built by the biggest AI companies, like Meta and xAI.
But whether or not a chatbot is built around engagement, none of them are required to stop and get a kid real help when the conversation turns dangerous.
Your Kids' First Love: A Chatbot
What are our kids using AI for? Parents picture homework help. But kids reach for something else.
- Children's chats with AI companion apps were nearly three times as likely to be sexual or romantic as about homework.8
- Nearly 1 in 5 high schoolers say they or a friend used AI for a romantic relationship.9
- This isn't confined to obscure apps – Meta and xAI build the same kind of companion chatbot too.12
No safety test. No age check. No parent in the loop. Common Sense Media, working with Stanford, rated AI companion apps "Unacceptable" for anyone under 18.14
Fake Nudes of Teenagers
It's not just companion chatbots. A fifteen-year-old girl in New Jersey found out classmates had used a free app to make fake nude images of her and 30 other students. She went public – and helped win the first federal law against AI deepfakes.
She's not alone: fake nude images have hit nearly 90 schools and more than 600 students worldwide since 2023, and the reporters tracking it call that an undercount. More on AI fakes and scams15
Life-Threatening Harm
These products reach kids without guardrails, and that's already threatening the safety of our kids – not hypothetically. After months with a Character.AI chatbot built to win his trust, a 14-year-old American died by suicide. Character.AI and Google settled with his mother and four other families in January 2026.2
No company admitted fault.
The American Psychological Association has issued two health advisories on AI and teens in 2025. Its CEO has warned the harm can be "life-threatening" – and in September, APA testified before Congress, asking lawmakers to act.1316
Where are the rules?
Medicine gets years of FDA review before it reaches your child. An AI chatbot that will talk to your kid about suicide gets none – tested on launch day, on your child.
The companies write their own child-safety rules, and the lobbyists keeping it that way cost about $226,000 a day.3
Meta's own AI rules · leaked, Reuters 2025
“Every inch of you is a masterpiece – a treasure I cherish deeply.”
It's bigger than one app
A company that can't stop its chatbot from grooming a 14-year-old is asking you to trust it with the most powerful technology ever created.
These companies aren't building a better chatbot – their own mission statements say the goal is AI that does everything a human can do. If they can't keep a chatbot safe for a child today, that is the warning, not a footnote.
The rules that catch this today are the rules that keep AI accountable as it gets more powerful.
The bigger picture
Keep humans in control
What you can do
We believe companies should be accountable for the harms their products cause.
87% of voters agree: AI companies should be liable when their technology is responsible for a child's suicide. Americans have won similar hard battles:4
- Parents and pediatricians got all 50 states to require car seats – cutting a baby's risk of dying in a crash by 71%.
- After a drunk driver killed her daughter, one mother built MADD in 1980 – and won a national law that saved nearly 32,000 lives.
We can require AI products to be checked before they reach a child too. See the standards we're asking for.
Questions people ask
Some AI tools help kids learn. But chatbots built for adults, companion apps, and "nudify" tools reach your child with no required safety testing first – and that's already threatening the safety of kids today. xAI's sexualized "Ani" companion, for one, sat in the App Store rated for ages 12+. For those products, companies still answer that question themselves – and that's the problem.
The real question is "who checked whether this is safe for a 13-year-old?" Right now, nobody is required to. Keep chatbots out of the bedroom at night, talk about what your kid uses, and know a chatbot is built to keep talking – not to keep your child safe. Common Sense Media's own guidance is blunt: no AI companion apps for anyone under 18.
One – and it shows this works. After families like that New Jersey teenager's pushed, Congress passed the TAKE IT DOWN Act in 2025: it makes AI deepfake nudes illegal and forces platforms to remove them within 48 hours. It's being enforced, with the first conviction in 2026. But it stops at fake images. No law makes a company safety-test a chatbot before it talks to your kid, and none makes them answer when one pushes a child toward suicide. That's the gap we're working to close.1011
More resources
These are independent resources – not from us. Each one is worth your time.
Common Sense Media: AI – Parent guides and research on how kids actually use AI, and how to keep them safe.
Design It For Us – A youth-led coalition pushing tech companies for safer products by design.
Sources20 references · tap to expand
Check everything – every number on this page comes from somewhere you can read.
- 1 ↩TIME · Raine v. OpenAI; 377-flag detail · October 2025
- 2 ↩CNN · Character.AI / Google settlement · January 7, 2026
- 3 ↩Issue One / Fortune · $226K/day AI lobbying spend · April 23, 2026
- 4 ↩ABF / OnMessage · 87% liability for child suicide · March 2026
- 5 ↩Pew Research Center · 64% of teens use AI chatbots · February 2026
- 6 ↩NBC News · Character.AI lawsuit; Sewell Setzer · October 2024
- 7 ↩NBC News · OpenAI blamed Adam's "misuse" · November 25, 2025
- 8 ↩Aura / PR Newswire · Kids' AI chats ~3x sexual vs homework · September 2025
- 9 ↩Center for Democracy & Technology / NPR · 1 in 5 high schoolers romantic AI · October 2025
- 10 ↩Congress.gov (S.146) · TAKE IT DOWN Act: deepfake NCII law · May 19, 2025
- 11 ↩TIME · First TAKE IT DOWN Act conviction · May 22, 2026
- 12 ↩Reuters / TechCrunch · Meta AI rules allowed romantic child chats · August 14, 2025
- 13 ↩APA (American Psychological Association) · APA CEO Arthur C. Evans Jr.: AI chatbot harm to teens 'can be life-threatening' · November 13, 2025
- 14 ↩Common Sense Media / Stanford · AI companion apps rated "Unacceptable" for under-18 · April 2025
- 15 ↩WIRED / Indicator · Fake nude images hit ~90 schools, 600+ students worldwide since 2023 · April 2026
- 16 ↩APA / Senate Judiciary Committee · Dr. Mitchell Prinstein's testimony urging Congress to regulate AI chatbots · September 16, 2025
- 17 ↩Dario Amodei, “The Urgency of Interpretability” · Amodei: we don’t understand how our own AI works · April 24, 2025
- 18 ↩CBS News, 60 Minutes · Pichai: AI is a “black box” we don’t fully understand · April 16, 2023
- 19 ↩CNBC (Musk at SXSW) · Musk: AI more dangerous than nuclear warheads · March 13, 2018
- 20 ↩Business Insider; Future of Life Institute · Altman: AI may “lead to the end of the world” · June 2015
Ready to demand commonsense rules for AI?
See what we're asking for →