Eight states sue OpenAI over teen self-harm. Demands: data deletion, AI test limits

Eight state attorneys general are taking OpenAI to court, alleging that ChatGPT encouraged some users, including teenagers, to hurt themselves or others, and that the company misled consumers about the tool's safety. The bipartisan group of AGs filed separate lawsuits Wednesday, and their demands are blunt: delete user data, verify ages, cap time on the chatbot, and stop testing advanced models outside secure, independently verified environments.

The states (Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, and West Virginia) are seeking penalties that range from $2,000 per violation in Nebraska to $50,000 per violation of that state's age-appropriate design law. Iowa wants up to $40,000 per consumer fraud violation, plus $5,000 for violations against people 60 or older. Arizona is asking for up to $10,000 per willful violation. The lawsuits also demand structural changes: Arkansas wants to halt further advanced AI testing and the creation of models capable of recursive self-improvement until safeguards are in place, and several states want advanced AI testing confined to secure, independently verified environments.

The core allegation is that OpenAI designed ChatGPT to win trust and keep users engaged, then failed to protect vulnerable users. Arizona Attorney General Kris Mayes alleges OpenAI used excessive agreement and simulated compassion to win user trust and collect personal data. 'OpenAI designed a product to attract young users and keep them engaged for as long as possible so it could capture their extensive personal data for the company's commercial benefit,' Mayes said. 'OpenAI put Arizonans safety at risk by failing to implement adequate user protections or transparency measures,' she added. Arkansas AG Tim Griffin put it more starkly: 'OpenAI assures us that ChatGPT is safe. The consequences of that assurance are lives lost and people harmed.' Idaho AG Raúl Labrador said, 'The safety of Idaho children cannot be an afterthought to runaway profit or technological advancements.'

The legal theory varies by state. The complaints accuse OpenAI of designing the chatbot to cultivate trust before reinforcing delusions or facilitating dangerous conduct. Idaho AG Raúl Labrador alleges OpenAI misled consumers, encouraged self-harm and violence, reinforced delusions, and offered dangerous medical advice. Kansas AG Kris Kobach says OpenAI designed features that encourage compulsive use, keeping consumers 'ignorant of the true nature of the product.' Montana AG Austin Knudsen alleges violations of privacy protections, including collecting children's data without required consent. Nebraska AG Mike Hilgers says OpenAI failed to protect children from harmful content and emotional dependency. West Virginia AG John McCuskey frames it as a public nuisance that endangers children. Arkansas AG Tim Griffin alleges OpenAI concealed known risks while assuring consumers that ChatGPT was safe, and unlawfully collected teens' personal information. Iowa alleges OpenAI misled consumers and concealed risks associated with autonomous AI agents.

The remedies are as notable as the allegations. Several states want OpenAI to delete data collected before a court judgment. Iowa specifically asks the court to order OpenAI to 'delete and destroy all Iowans' data obtained prior to the entry of this Court's judgment.' Nebraska wants 'accessible and easy-to-use options to limit the amount of time' users spend on ChatGPT, plus parental controls and age verification. West Virginia wants default session and nighttime limits for children, mandatory crisis-escalation protocols, an independent monitor, and funding for youth crisis-response capacity and digital-safety education in public schools. West Virginia also seeks $5,000 per repeated consumer protection violation, along with deletion and cessation of use of personal data it alleges was unlawfully collected from West Virginia minors. Iowa is also asking the court to halt further testing of advanced AI models until safeguards are independently verified to not be posing a danger to the public and until third-party monitors have verified the same.

Other states are pushing for similar or stricter terms. Idaho seeks $5,000 per consumer protection violation and court-ordered safeguards, including age verification and limits on harmful or addictive features. Kansas is seeking $10,000 per consumer protection violation and $20,000 per violation involving a protected consumer, plus restitution and a permanent injunction against deceptive and unconscionable practices. Montana wants penalties of $10,000 per consumer-protection violation and $7,500 per privacy violation, deletion of unlawfully obtained personal information, and notification to the attorney general's office should an AI agent go rogue again. Arizona also wants to bar OpenAI from deploying or testing advanced models outside a secure, independently verified environment and require prompt reporting of incidents in which models escape controls or access systems without authorization. Arkansas is seeking financial penalties and an injunction against collecting data from teens.

This is not a single coordinated lawsuit but eight separate filings from a bipartisan group of attorneys general, which signals that AI safety has become a state-level enforcement priority. The demands go beyond fines: they target how OpenAI develops and tests models, including restrictions on advanced AI testing and, in Arkansas's case, recursive self-improvement. For any company deploying generative AI to consumers, the message is that state regulators are watching how models interact with minors and are prepared to demand design changes, not just write checks.

The stakes for OpenAI are immediate. The company now faces eight separate state actions, and the remedies could reshape how it handles user data, model testing, and product design. For founders and operators building AI products, the practical takeaway is that safety claims are now legally testable: if you tell users a tool is safe, states may hold you to that promise in court. The specific demands (age checks, time limits, crisis protocols, independent verification of advanced models) show what state regulators expect from consumer AI, and they set a marker for the industry.