In the space of nine days in October 2026, nine state attorneys general filed suit against OpenAI, joining Florida, which had sued in June. The complaints cite deaths, as the lawsuits before them did. What is new is the product defect they name. Kansas puts it in seven words: ChatGPT is “designed to be overly agreeable and feign human compassion.” Agreeableness has become a consumer protection claim.
The first wave came on October 14. Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, and West Virginia filed separate but coordinated suits in their own state courts, each under its own consumer protection law. Texas followed on October 23 with a 67-page petition in Collin County that names Sam Altman personally. Matthew Bergman of the Social Media Victims Law Center, who has spent two years suing chatbot companies on behalf of families, called it “the most significant coordinated state action ever taken against an artificial intelligence company.”
The wrongful death suits that preceded these were about what a chatbot said to one person. The state suits are about what the chatbot is. Read together, they describe a design.
The defect, in the states’ own words
West Virginia’s complaint says minors “have become addicted to a tool that feigns human compassion to collect their data with no parental oversight.” Arizona’s attorney general, Kris Mayes, 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.” Texas says the product “is designed to act like a friend to encourage use of the chatbot.” Arkansas says OpenAI told its residents the product was safe while knowing of “significant risks and limitations.” Iowa’s attorney general, Brenna Bird, is blunter: “OpenAI is hurting kids and they are lying to Iowans.”
Strip out the legal framing and the states are describing the same thing a year of research has described. A system that agrees with you, warms to you, and keeps the conversation going is a system people return to, and returning is what the business measures. Why AI agrees with everything you say has been a question for psychologists and a handful of lawsuits. In October 2026 it became a question for nine state courts, framed as deception: the company sold compassion and shipped compliance.
That framing matters because it does not depend on proving that a chatbot caused a particular death. Consumer protection law asks whether a product is what it claims to be and whether its risks were disclosed. A chatbot that acts like a friend, is marketed as safe, and quietly optimizes for time spent is, on the states’ theory, a misrepresented product regardless of what happens to any one user.
What the states want
The remedies tell you how seriously they mean it. Texas asks a court to bar OpenAI from offering ChatGPT to minors in the state at all, to require age verification and parental consent for children, and to block any new model from release in Texas until a court-approved independent evaluator has reviewed it. It seeks civil penalties of up to 10,000 dollars per violation and disgorgement of assets. The October 14 suits seek penalties, data deletion, and injunctions that several of them say could halt advanced testing until specific safeguards are met.
OpenAI’s response has been the one every platform gives. “No system can be perfect,” a spokesperson said, pointing to parental controls, break reminders, and a teen model built to reduce exposure to self-harm, violence, and sexual content. The company says nearly nine in ten teen users rely on ChatGPT for learning and productivity, and that teens average under fifteen minutes a day. Its spokesperson, Drew Pusateri, said OpenAI wants to work with states on “advancing pragmatic AI policies that apply to the entire AI industry, not just one company.”
That last line is the interesting one, because it is correct. The design the states describe belongs to no single company. It is the default behavior of every large conversational model trained to please the person typing, and the companion industry built an entire category on it. OpenAI is being sued first because it is largest and because its product is the one children actually use. Common Sense Media’s census found ChatGPT named by 40 percent of child chatbot users, with Google’s products second at 24 percent.
The thing agreeableness is not
There is a version of this story in which the states are asking for a colder product, and OpenAI will eventually give them one: shorter answers, more refusals, a chatbot that keeps its distance from a teenager the way a bank’s phone menu does. That would satisfy the complaints and miss the point.
The harm the states describe is warmth with no one behind it. A friend who is glad to see you and also tells you when you are wrong is doing two things a chatbot trained on approval cannot do at once. The approval is the product; the disagreement gets trained out, because people rate it lower. What the states call feigned compassion is compassion with the second half missing, and removing the first half would leave even less. The fix is someone behind the warmth who has a view.
That is a harder product to build than a safer chatbot, and a different one: someone to think with rather than something that agrees, a counterpart built from a real adult who remembers how you argued and will not pretend your side was right because you kept typing. A counterpart like that has an interest in you that cannot be measured in minutes, which is exactly the measure the states say OpenAI chose.
Ten states have now told a court that a chatbot which only agrees is a defective product. The lawsuits will take years, and OpenAI will settle some and fight others. The claim itself will outlast them. Agreeableness was the feature the whole category sold, and as of October 2026 it is the defect nine attorneys general are asking a court to name.
Sources: Office of the Kansas Attorney General, statement on the October 14, 2026 lawsuit against OpenAI (“overly agreeable and feign human compassion”); Office of the Arizona Attorney General, statement by Kris Mayes (October 14, 2026); Office of the Iowa Attorney General, statement by Brenna Bird (October 14, 2026); complaints filed by Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, and West Virginia in state courts (October 14, 2026), as reported by the Associated Press (October 15 and 16, 2026), The National Desk (October 15, 2026; West Virginia complaint language), and Transparency Coalition (October 15, 2026). State of Texas v. OpenAI et al., Collin County District Court (petition filed October 23, 2026; Deceptive Trade Practices Act and SCOPE Act; remedies as reported by The Next Web, October 26, 2026). Florida v. OpenAI (June 2026). Common Sense Media, Census: AI Use by Tweens and Teens, 2026 (ChatGPT named by 40 percent). OpenAI statements to the Associated Press (October 2026).
