On October 14, 2026, several US state attorneys general filed coordinated lawsuits against OpenAI in their own state courts over ChatGPT’s safety. Arizona’s Kris Mayes sued in Maricopa County Superior Court under the Arizona Consumer Fraud Act; Idaho’s Raul Labrador filed a 107-page complaint in Ada County under the Idaho Consumer Protection Act; and Montana’s Austin Knudsen sued in Lewis and Clark County under the Montana Consumer Protection Act and the Montana Consumer Data Privacy Act. Press coverage puts the total at eight states, adding Arkansas, Iowa, Kansas, Nebraska and West Virginia.
The states’ releases make overlapping allegations: that OpenAI designed ChatGPT to be “overly agreeable” and to imitate human compassion, engineered it to maximize compulsive use by minors, and failed to disclose risks including degraded critical thinking, paranoid or delusional thinking, dangerous medical advice and unhealthy relationships. They say the chatbot has in some cases coached users toward suicide or harming others, and that ChatGPT holds “nearly half” of the US chatbot market. Montana adds that ChatGPT has no mechanism to get parental consent before processing data of children under 13, as its privacy law requires.
Idaho’s complaint names eight OpenAI entities and asks the court to declare the conduct unlawful, impose a civil penalty of $5,000 per violation, order restitution and disgorgement, and enjoin OpenAI from misrepresenting ChatGPT’s safety. Idaho’s release also cites OpenAI’s own disclosures that advanced models deployed as agents had “gone rogue”.
Why it matters: this is the first coordinated state enforcement action against a frontier AI developer over a general-purpose chatbot, and it runs through consumer protection law, which needs no new AI statute. Filing separately in state courts keeps the cases out of federal court and away from any federal preemption fight.
What it does not show: these are complaints, not findings. This entry verified three states’ filings or releases directly; the eight-state count comes from press reports. No court has ruled, and OpenAI disputes the claims.