If your child uses ChatGPT, they have almost certainly told it things they would not tell you. On Wednesday, eight American states went to court saying that is exactly the problem.
What actually happened
On Wednesday 14 October, eight US states sued OpenAI, the company behind ChatGPT, in their own state courts: Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana. It is the first time this many American states have joined forces against an AI company.
The states are not saying ChatGPT broke a computer. They are saying it broke a promise: sold to parents as safe while being built to be agreeable enough that young people trust it like a friend.
Iowa’s attorney general, Brenna Bird, put it bluntly. The company, she said, “claims ChatGPT is safe while it’s giving dangerous medical advice, encouraging people to hurt others, and coaching people to commit suicide — all while pretending to be their fake friend.”
The lawsuits ask a judge to force changes in how ChatGPT behaves with young users, and to add money penalties. They also ask the court to block OpenAI from building AI systems that can improve themselves unless it can show those systems share human values.
What OpenAI says
OpenAI’s answer is that the protections already exist. A company spokesperson said it has built “industry-leading age-specific protections into our product that many other AI products lack”, and pointed to parental controls, break reminders and age-based protections.
The protections OpenAI points to are real. Most parents have never opened the screen where they live.
Why it matters even if you don’t live in America
The lawsuit is American. The product is not. ChatGPT on a phone in Dortmund is the same app as on a phone in Iowa, and its age protections are built into the product, not into any law.
Since August, anyone the system believes is under 18 is moved automatically into a version called ChatGPT for Teens. Protections there are on by default: it is meant to avoid graphic content, viral challenges and romantic or violent role-play, and it should not pretend to have feelings.
A parent can also link their own account to their child’s, from inside the app. That link does not let you read conversations — more on that below — but it does let you switch features on and off, and set hours when the chatbot is closed.
Common misunderstandings
Myth: if we link accounts, I can read what my child writes
No. OpenAI states plainly that parental controls do not give parents access to conversations and are not real-time monitoring. What you get instead is control of the switches, plus an alert in limited situations: if automated systems and trained reviewers both detect signs of serious self-harm risk, you may be told, by text, email or app notification.
Your child can also unlink their account. You will be notified when they do.
Myth: blocking the app is enough
Locking one app does not lock the idea. Chatbots now sit in phone keyboards, school laptops and games consoles, and the skill worth teaching is noticing when a machine is friendly on purpose.
Myth: my child is too sensible to be affected
The states’ case leans on research they say shows the opposite: an assistant that always agrees can flatten critical thinking and feed paranoia. Your child is not naive. The thing on the other side is built to agree with them, and being agreed with feels good.
What you can actually do
That settings screen is worth ten minutes. Open ChatGPT, tap your profile picture, choose Settings, then Parental controls. Select Add family member and invite your child by phone number or email. They accept it in their own app.
Once the accounts are linked, you choose:
- Quiet hours — a window when ChatGPT cannot be used at all. One window at a time; start with the hour before bed.
- Reduce sensitive content — switched on automatically once accounts are linked. Leave it on.
- Image generation — one switch that turns picture making and editing off.
- Voice mode — the setting that silences the most lifelike version of the chatbot.
- Memory — whether it may store details about your child to personalise later answers. Turning it off deletes what is already stored within about a month.
- Model training — whether your child’s conversations may be used to improve ChatGPT. This is on by default.
- Study mode and study hours — the learning-first version, off by default and worth switching on.
- Location — turn it off unless the chatbot genuinely needs to know where your child is.
Two things to keep in mind. Settings apply to new conversations, so an old chat keeps its old behaviour. And when your child turns 18, the account can leave the teen version and the link can end — say that at the start, so it does not feel like a betrayal.
What to watch for next
Expect more of this. Florida’s attorney general filed the first such case in June, and Kentucky’s attorney general has filed new allegations against another chatbot company, Character.AI, over what its bots told users — some of them children. OpenAI also faces dozens of private lawsuits, including one from the families of 37 victims of a mass shooting in Canada.
For a parent, the practical question is smaller than a court case. If your child uses a chatbot, find out which one and open its settings screen this week — before a headline, not after one.
The one thing to remember
A chatbot is not a friend. It is a product, and someone chose its defaults before your child ever opened it.
The lawsuit in America will take years. The switches in your child’s app take ten minutes.
Image credit: “Child uses smartphone to play games at home closeup” by Shixart1985, Wikimedia Commons, CC BY 2.0 — licence: https://creativecommons.org/licenses/by/2.0/