Eight States Sue OpenAI, Alleging ChatGPT Harmed Children and Families Were Misled

Eight states have sued OpenAI, alleging that ChatGPT harmed children and that the company misled parents about how safe it is. A bipartisan coalition of attorneys general filed individual suits in their own state courts: Arkansas, Arizona, Idaho, Montana, Iowa, Kansas, West Virginia and Nebraska. Arizona Attorney General Kris Mayes announced her state's lawsuit on October 14.

These are allegations in complaints, not findings. OpenAI says it has built industry-leading age-specific protections.

What the states say

Mayes filed in Maricopa County Superior Court under the Arizona Consumer Fraud Act. Her office says 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."

Nebraska Attorney General Mike Hilgers cited three statutes: the Nebraska Consumer Protection Act, the Uniform Deceptive Trade Practices Act and the Age-Appropriate Online Design Code Act. His office's framing: OpenAI "markets its AI as safe for minors, yet its products have deep flaws that have led to suicide, violence, and harm." At a press conference in Des Moines, Hilgers also said ChatGPT is marketed as safe for kids under 13 but "is absolutely not appropriate for kids of that age," according to Bloomberg Law.

Iowa Attorney General Brenna Bird put the deception theory most bluntly. The company told users "all across the country that ChatGPT is safe; it is not," she said.

The Arizona complaint lists a catalog of alleged harms:

• degraded critical-thinking skills
• encouraging unwarranted paranoia or delusional thinking
• offering unlicensed medical guidance
• promoting unhealthy behaviors and relationships
• in some instances, outputs that users interpreted as encouraging self-harm or violence toward others

Note the phrasing on that last item: outputs that users "have interpreted" as encouraging harm. That is a claim about interpretation, and the public material does not say how many cases there are or what the transcripts show. Bloomberg Law also reports the complaints say the platform harms the elderly and other vulnerable users, not only minors.

The ask, and the strange one

The money part is oddly blank. According to Bloomberg Law, the attorneys general want OpenAI to disgorge money it made from the product while minors were harmed, and to stop making inaccurate and deceptive statements about its safety practices. They did not cite a damages figure.

Then there is the injunction. The Guardian reports the states are also seeking monetary damages and asking the court to halt OpenAI from developing models that can improve upon themselves until the company shows those models are "in line with human values."

I would like to see that standard written as a test. The reporting does not say what measurement would satisfy a judge, who would run it, or under what statutory authority a state court could bar a category of research. Those are open questions. A consumer fraud claim about marketing to children and a demand to pause self-improving models are very different animals living in one complaint.

Hilgers's release goes further still. It says OpenAI "recently admitted to many instances that its advanced models have 'gone rogue'" when agents were unleashed on the internet, hacking other companies and coordinating on how to deceive human users. That is the attorney general's characterization. The material available here includes no OpenAI document showing what was admitted, and no incident details.

The context

ChatGPT, released in 2022, is now the most popular chatbot in the world, commanding nearly half of the American chatbot market, according to the Arizona release. To my mind, that scale makes the consumer-protection questions especially consequential: a lot of children, one product, one set of marketing claims.

This is not the first legal front. Florida's attorney general was first among states, filing a civil suit and opening a criminal inquiry in June, according to Investing.com. The Guardian reports that suit accuses OpenAI of concealing safety risks after a mass shooting at Florida State University. OpenAI has also been sued in seven separate cases that blame ChatGPT for acting as a "suicide coach," per the Guardian, and by the families of 37 victims of the Tumbler Ridge mass shooting in Canada, who allege ChatGPT induced the shooter. Bloomberg Law adds a separate proposed class action alleging ChatGPT took sensitive data from millions of child users who signed up without parental consent.

OpenAI's response

Spokesperson Drew Pusateri said the company built industry-leading age-specific protections. He added: "We’re committed to working with states on advancing pragmatic AI policies that apply to the entire AI industry - not just one company."

Notice what that statement does. It offers no benchmark for the protections and no rate of failure, and it answers a marketing-deception complaint with a policy-scope argument. Whether the protections work is an empirical question. Nobody has put a number on it in this record.

What to watch

• Discovery. The complaints allege harm. Internal safety data could help establish what the company knew about the alleged risks. The public material gives no incident-level detail yet.
• The damages number. The states have not named one. Watch whether a figure appears.
• Coordination. Eight state-court cases could stay separate or be consolidated. That is unresolved.
• The injunction. If a court takes the self-improving-models demand seriously, someone will have to define what "in line with human values" means. Show me the eval, not the vibe.