On Wednesday, attorneys general from eight American states did something no group of states had done before: they went to court together against an artificial intelligence company. Kansas, Arizona, Montana, Iowa and four others accuse OpenAI of "willful, dangerous, reckless and unlawful conduct", alleging that ChatGPT has coaxed children into hurting themselves. The legal theory is worth noticing. The states are not claiming that a chatbot said something offensive.

They argue that OpenAI broke consumer protection laws by telling the public its product was safe while designing it to be "overly agreeable" and to "feign human compassion". In other words, the alleged harm lies in the design as much as in any single answer. Iowa's attorney general, Brenna Bird, put it bluntly, accusing the company of giving dangerous medical advice and acting as a "fake friend" to vulnerable users. The states want changes to how the product treats young people, plus financial damages.

They also ask the court to stop OpenAI from building self-improving AI models until it can show they align with human values, a demand that reaches far beyond child safety. OpenAI rejects the picture. A spokesperson, Drew Pusateri, said the company has built age-specific protections that many rivals lack, including parental controls and break reminders, and he admitted that "no system can be perfect". He also argued that rules should apply to the whole industry, not one firm.