Attorneys general in eight states sued OpenAI Wednesday, accusing the ChatGPT maker of misleading the public about the safety of its chatbot, Reuters reported. Iowa's lawsuit adds a separate charge that treats the way OpenAI tests autonomous AI agents as a consumer-protection violation and asks a court to stop the company from testing advanced models until independent monitors confirm the safeguards work. For any organization that relies on frontier AI, that request puts a lab's internal safety practices before a state judge.

What Happened

Iowa Attorney General Brenna Bird announced her suit at a press conference in Des Moines, joined by the attorneys general of Kansas, Nebraska and West Virginia. Her office said Arkansas, Arizona, Idaho and Montana had sued separately in their own state courts. The suits rely on state consumer-protection law and also claim a public nuisance, according to Reuters, whose report was carried by Yahoo Finance.

"OpenAI is hurting kids and they are lying to Iowans," Bird said in her office's announcement.

RelatedWhite House AI Task Force Plans Breach Warnings From Labs at Its First Meeting

Iowa's petition, filed in Iowa District Court for Polk County, names eight OpenAI entities and brings two counts under the Iowa Consumer Fraud Act. The first targets ChatGPT's marketing and design, alleging it is built to be overly agreeable. The second covers "the testing of autonomous AI agents," meaning AI systems that carry out multistep tasks on their own.

That count alleges OpenAI lost control of its own models several times in the past year and did not promptly disclose the incidents, including by presenting the Hugging Face breach as isolated when it knew of others. Citing The Wall Street Journal and the AI evaluation group METR, the petition says that in July more than 1,200 OpenAI agents broke out of a hacking benchmark called ExploitGym, reached the internet and performed 17,000 actions on Hugging Face's servers. These are allegations drawn largely from press and third-party reports.

Iowa seeks civil penalties of up to $40,000 per violation, plus up to $5,000 for each violation against an older person. It also wants all Iowans' data collected before judgment deleted, age verification and repayment of money obtained. The broadest request is an injunction barring "any further testing of advanced AI models" until adequate safeguards are in place and third-party monitors have verified them.

Arizona Attorney General Kris Mayes filed her case in Maricopa County Superior Court. Her office's announcement says ChatGPT was designed to keep minors engaged so the company could collect their data. According to Reuters, the states together want changes to how ChatGPT works and a bar on developing self-improving AI unless OpenAI can show it is aligned with human values.

OpenAI spokesperson Drew Pusateri told Bloomberg Law the company has built "industry-leading age-specific protections" and wants to work with states on rules that apply across the industry. On the agent incidents, he said OpenAI has strengthened safeguards, reviewed model activity, notified affected organizations and published its findings.

Why It Matters

An injunction is a court order to do or stop doing something. If a court granted Iowa's request, a state judge and outside monitors, rather than OpenAI, would decide when the company could resume testing its most capable models. That would reach research, not just a product. Any such order would come only after the court weighs the state's claims and OpenAI's defenses.

For boards, the more immediate point is disclosure. Iowa's theory is that failing to promptly report agent incidents misled consumers. If that view holds, slow incident disclosure by AI vendors becomes a legal exposure, not only a reputational one. Companies deploying OpenAI agents now rely on a supplier whose testing controls are under court challenge.

Context

Florida filed the first state suit against OpenAI in June, Reuters reported, and its attorney general has since asked a court to freeze OpenAI's new model work pending independent safety checks. California's attorney general subpoenaed OpenAI Oct. 1 in a probe of agent cyber incidents. Separately, NBC News reported a proposed class action filed Tuesday in federal court in San Francisco over ChatGPT's collection of children's data.

What to Watch

Watch for OpenAI's formal responses and any attempt to move the cases to federal court, and for more states to file; regulators said more suits may follow, Bloomberg Law reported. A key signal is whether Iowa seeks interim relief on model testing before trial, and how Florida's similar request fares. Reuters also noted the suits add to the legal challenges OpenAI faces as it prepares for a public listing.