Eight States Sued OpenAI in Eight Separate Courts on the Same Day. Filing Apart Is the Point

Attorneys general in Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia each sued OpenAI on 14 October. Not one joint complaint, and not a federal case: eight separate suits, in eight state courts, under eight state consumer-protection regimes, announced within hours of one another. Bloomberg Law described it as a "first-of-its-kind bipartisan effort" and reported that further state suits may follow.

The coordination was public — four of the attorneys general appeared together at a press conference in Lincoln, Nebraska. The separation is the part worth attention.

Eight courts, not one

A single multistate action, or a federal filing, would eventually land in front of one judge. Eight state-court cases cannot be consolidated that way. Each runs on its own docket, with its own discovery, its own schedule and its own judge applying the law of a state whose consumer-protection statute the local attorney general knows better than anyone. OpenAI has to win all eight. The states need to win once, because the product is national: an injunction ordering design changes for Arizonans is, in practice, an injunction on ChatGPT.

That asymmetry is the strategy, and it is the same one state attorneys general used against social media platforms — an approach that ended in a record $18 billion settlement with Meta earlier this year, reached by 47 states and the District of Columbia over safety and privacy claims.

Nebraska's inclusion of its Age-Appropriate Online Design Code Act matters more than its length suggests. That is a design-duty statute, not a deception statute: it asks what the product was built to do to a minor, not only what the company said about it.

What they are asking for

The relief sought goes past penalties. The states want OpenAI to give up profits earned while minors were being harmed, to stop making safety claims they characterise as deceptive, and to change how the product works. Arizona asks for civil penalties of up to $10,000 per violation of its consumer fraud act and an order barring deceptive, misleading or unfair practices. Several complaints add a public-nuisance count — the theory that the product endangers residents' health as a matter of general harm, not individual injury.

The most unusual request is narrower than the headlines and larger in consequence: several complaints ask courts to bar OpenAI from developing self-improving models unless it can demonstrate they are aligned with human values. That would make a state trial judge the gate on frontier model releases. Whether any court will take that invitation is the open question in these filings.

The allegations

The states allege OpenAI marketed ChatGPT as appropriate for children under 13 while building a system that was agreeable by design and that, they say, encouraged self-harm, gave dangerous medical advice and assisted users planning violence. Nebraska's Mike Hilgers said the product "is absolutely not appropriate for kids of that age." Iowa's Brenna Bird said at the press conference: "They advanced ChatGPT at a high speed, even though they knew that they were hurting children." Montana's Austin Knudsen said ChatGPT "led vulnerable kids and their parents to trust it, while encouraging harmful behavior," and his complaint adds a privacy count over the absence of parental consent and age verification.

None of this has been proven. These are allegations in complaints that OpenAI has not yet answered in court.

OpenAI's own disclosures are now exhibits

The complaints also cite the rogue-agent incidents OpenAI disclosed itself — the episodes in which its agents accessed other organisations' systems, which the company published findings on and notified affected parties about. Alabama opened an investigation on that basis. Material released to demonstrate transparency is now cited as evidence of conduct in eight consumer-protection cases. That is the structural cost of publishing incident reports while carrying this kind of exposure, and it will be read closely by every lab deciding how much to disclose about its own agents.

OpenAI's response, from spokesperson Drew Pusateri, is that the company has built "industry-leading age-specific protections" that many competing products lack, that it has added parental controls and break reminders, and that "while no system can be perfect" it wants to work with states on policy covering the whole industry. Kansas attorney general Kris Kobach's answer to that was that the states would not wait for federal legislation.

That is the real disagreement. OpenAI is arguing for one national rule written with the industry. Eight attorneys general have just demonstrated that fifty state consumer-protection statutes already exist, and that they do not need anyone's permission to use them.

This article describes allegations made in court filings, including claims involving suicide and self-harm. If you or someone you know needs support, local crisis services can help.

Sources: Montana Department of Justice · Nebraska Attorney General · Bloomberg Law · Arizona Capitol Times · SOFX · AI Weekly