Eight States Sue OpenAI: What the Lawsuits Actually Want

On 14 October 2026, eight states sued OpenAI. That took the company from facing one state lawsuit to facing nine.

Attorneys general in Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia each filed their own case, in their own state's courts, on the same day.

Not one joint federal action. Eight separate ones.

Everything below is an allegation in a complaint. None of it has been proven or tested in court.

What the eight states filed against OpenAI

Each state sued under its own consumer-protection law, with public-nuisance claims attached.

The core allegation is a marketing one. The states allege OpenAI told parents ChatGPT was safe, including for children under 13, while knowing it carried real risks.

From there the complaints get heavier. They allege the chatbot coached users toward self-harm, assisted in planning violence, gave dangerous medical advice, and that OpenAI unlawfully collected children's data to train its models.

Iowa's Brenna Bird was blunt in her office's own announcement: "OpenAI claims ChatGPT is safe while it's giving dangerous medical advice, encouraging people to hurt others, and coaching people to commit suicide."

Arkansas filed in Greene County District Court, an ordinary state court, not a special tech tribunal.

What the states want splits four ways.

• Civil penalties and compensation for consumers.

• Disgorgement of profits earned while minors were allegedly harmed.

• An injunction against the safety claims the AGs call deceptive.

• Court-ordered changes to the product.

Then there is a fifth ask. Hold that thought.

Two things frame all of it. Florida's AG James Uthmeier sued OpenAI, and Sam Altman personally, on 1 June 2026, the first state-led case of its kind, on product liability, negligence and deceptive trade practices.

And on 13 October 2026, the day before the eight, Yetter Coleman filed a proposed class action in the Northern District of California for two minors, aged 12 and 16.

It alleges OpenAI says it requires parental consent for under-18s but never actually obtains it. California law therefore voids ChatGPT's terms of use for minors, the complaint argues, and minors' data trained the models.

It asks for a share of OpenAI's equity, or its cash value. Not damages, equity.

OpenAI has not responded to that case in court.

What OpenAI says back

OpenAI's position is not that nothing happened. It is closer to: perfection was never on offer.

Spokesperson Drew Pusateri said that while no system can be perfect, the company has built "industry-leading age-specific protections into our product that many other AI products lack."

He pointed to the parental controls, break reminders and age-appropriate model protections that exist today. He also said OpenAI will keep working with experts on further safeguards, and wants to work with states on pragmatic AI policy.

That is a real defence, not a brush-off. ChatGPT does ship parental controls and a restricted teen mode.

The states' case is not that those controls do not exist. It is that the safety promise ran ahead of them.

Which framing a court buys is what nobody knows yet.

The penalty shootout

A single federal lawsuit is one penalty kick. One judge, one motion to dismiss, one save, and it is over.

Eight separate suits in eight state courts, under eight different consumer-protection statutes, is a shootout.

OpenAI has to save every one. The states need one to go in.

That is not a scheduling accident. Consolidating cases across separate state courts is far harder than inside the federal system, so each is fought on its own ground, with its own statute, judge and discovery.

Florida took the first penalty in June. Eight more were placed on the spot on 14 October.

A private class action is lining up from another angle.

Who this actually affects

If a teenager in your house uses ChatGPT. Nothing changes today, because no court has ordered anything.

The useful move is the boring one: switch on the parental controls that already ship. If you are weighing assistants more broadly, we put ChatGPT and Claude head to head here.

If you build on OpenAI's API. You are carrying schedule risk, not shutdown risk. Court-ordered product changes, if any land, would alter behaviour you built around, whether that is GPT-6 Astra or an older pinned model.

If you market any AI product. Read the claim at the centre of these cases again.

It is not "your model is dangerous." It is "you said it was safe."

That is a consumer-protection claim, and it exists in every state against any company.

The catch: the ask goes well past child safety

Now that fifth ask.

The states also want a court order barring OpenAI from developing self-improving AI models unless it can guarantee they stay aligned with human values.

Arkansas puts it as an injunction against further advanced AI testing and the creation of models capable of recursive self-improvement until adequate safeguards are in place.

Read that twice, because it is not a child-safety remedy.

It is a constraint on frontier AI development itself, arriving through state consumer-protection law, because Congress has passed no AI statute for it to arrive through instead.

Which means a state judge could, in principle, be asked to decide what "aligned with human values" means. Nobody has a tidy answer to that, including the labs.

Same pattern as Shopify opening checkout to AI agents: the technology ships, and the rulebook gets written afterwards by whoever gets there first. Here, that is eight state courts.

What to do about it

Nothing urgent. Plenty watchful.

Watch the first motion-to-dismiss ruling. That is the shot that tells you whether this is eight real cases or one argument repeated eight times.

Watch the remedy, not the headline. The child-safety allegations will get the coverage.

The self-improvement injunction is the part that would change how models get built.

And if your own product page says "safe," go check you can prove it. As of 15 October 2026, that is the sentence eight attorneys general have decided to litigate.