Eight state attorneys general sued OpenAI on Wednesday, each in their own state's courts, arguing that the company marketed ChatGPT as a safe product while knowing what it was capable of telling the people using it.
The states are Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia. Every suit rests on state consumer protection law — the same body of law states use against deceptive advertising — rather than on any statute written for artificial intelligence, which is a measure of how little of that law exists.
The claim is about what OpenAI said, not only what the product did
The common allegation is deception: that OpenAI told users, and the parents of users, that ChatGPT was safe, and that the company's own knowledge of the risks did not match the claim.
Underneath it sits a design argument. The complaints describe a chatbot built to be agreeable and engaging in order to hold a user's attention, and argue that the same quality makes it dangerous when the user is a teenager, or in crisis, or planning harm. The states say the product reinforced users' delusions, failed to put adequate safeguards around conversations about suicide and self-harm, and in some cases gave information to people who went on to commit violence.
Iowa Attorney General Brenna Bird said the company is "hurting kids and they are lying to Iowans." West Virginia's JB McCuskey framed the suit around the absence of guardrails in conversations about suicide and school shootings: "We must do all we can to protect kids." Nebraska's Mike Hilgers put it more narrowly, saying the case "simply holds OpenAI to the same standards under Nebraska law as any other company."
What they want from a judge
Two things, and the second is the unusual one.
The first is an order forcing substantial changes to how ChatGPT operates — the ordinary shape of a consumer protection remedy.
The second asks a court to bar OpenAI from developing self-improving AI models until it can show they are aligned with human values. That is a request for a judge in a state court to impose a condition on frontier model research, and it has no real precedent. Whether any court will grant it is a separate question from whether the deception claims survive, and the two will likely move at different speeds.
OpenAI's answer
Spokesperson Drew Pusateri said the company has "built industry-leading age-specific protections into our product that many other AI products lack," pointing to parental controls, break reminders and age-appropriate model behavior. He acknowledged that no system is perfect and said more safeguards are coming. He also said OpenAI wants to work with states on rules that apply across the industry rather than to one company — an argument that these eight separate state suits, filed in eight separate courts, are exactly the wrong instrument.
The eight are not the first
Florida sued in June, and is not among this week's group. British Columbia filed against OpenAI in a California court in September. At the start of September, victims and families filed 30 suits in the Northern District of California. On Tuesday, a mother filed a proposed class action in San Francisco federal court arguing that OpenAI profits unlawfully from data that minors share with ChatGPT, since under California law minors cannot consent to that use.
The pattern worth watching is the shift in who is suing. Private plaintiffs have to prove harm to a particular person. An attorney general does not: a consumer protection case turns on whether the public was misled, which is a lower bar and a wider one. Eight of them filing in the same week puts OpenAI in front of eight different judges on substantially the same question.
Reporting from Al Jazeera, Just Security and The Hill.
