Eight states sued OpenAI in mid-October, accusing the company of harming children and misleading parents about the risks of ChatGPT. Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana each filed separate, coordinated lawsuits in their own state courts, according to a summary of NBC News reporting by the Business and Human Rights Centre. Florida had already filed its own suit against OpenAI earlier in the year.
Reuters reported the filings as they landed:
JUST IN: Eight US states sue OpenAI, claiming ChatGPT harms children https://t.co/FmzSxBc0GL pic.twitter.com/vQd30nM60y
— Reuters Legal (@ReutersLegal) October 14, 2026
The headline is the number of states. The more consequential detail is the legal theory they are using, because it does not depend on any new AI law at all.
What the states allege
The lawsuits rely on state consumer protection laws, the same statutes attorneys general use against misleading advertising and unfair business practices. The core claim is that OpenAI marketed ChatGPT as safe while designing it in ways that create risk. According to the Business and Human Rights Centre’s summary, the states argue ChatGPT’s design encourages excessive use and emotional dependency, and that the product has pushed users toward suicide or self-harm, helped with violence against others, and given dangerous medical advice.
Idaho’s attorney general, Raúl Labrador, put the design argument most sharply. His office described ChatGPT as designed to be overly agreeable, accommodating, and to feign human compassion, and said OpenAI misled the public about its safety, LEX18 reported. The states are seeking financial penalties and court-ordered safeguards for ChatGPT.
None of these claims has been tested in court.
Why the legal theory matters
There is no comprehensive federal law governing AI chatbots, and Congress has not passed one. These states found a way around that. Consumer protection law does not care whether a product is a chatbot, a toy or a supplement. It asks whether the company told customers the truth about it and whether its practices were unfair. If a court accepts that marketing a chatbot as safe while building it to be agreeable and emotionally engaging can be a deceptive practice, that reasoning would not stay limited to one company.
The target is also notable. These suits focus less on any single harmful answer than on how the product is designed to behave: agreeable, engaging, emotionally responsive. Those are features that make chatbots pleasant to use and keep people coming back. The states are arguing that, at least for children and vulnerable users, the same features are the hazard.
OpenAI’s response
OpenAI told NBC News that no system can be perfect and that it will keep working with experts, pointing to the parental controls, break reminders and age-appropriate model protections that already exist, according to the Business and Human Rights Centre. It also said it is committed to working with states on practical AI policies, and that those rules should apply to the entire AI industry, not just one company.
That last point is a fair one, and it doubles as an argument. Chatbots from many companies share the same basic design: trained to be helpful and agreeable, available at all hours, and able to sound warm. If the states’ theory is right, it is a problem for the category, not a single product. If it is wrong, eight states will have spent heavily to establish that being friendly is not a deceptive practice.
What to watch
The early rulings will matter more than the final outcomes. Watch whether courts let the design-based claims go forward, or narrow the cases to specific statements OpenAI made about safety. Watch whether more states join, and whether any settlement includes concrete product changes such as stronger age verification or limits on how emotionally responsive a chatbot can be with minors. Those terms, more than any penalty, would show what a safe chatbot is supposed to look like.
Related reading: AI Agents Sent Wikipedia Millions of Requests. Wikimedia Had to Investigate Whose They Were. · 153 Million Driver’s Licenses Leaked From a Company That Only Needed to Check Your Age
