Eight US states sue OpenAI, alleging ChatGPT harms children

Attorneys general in Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia filed separate lawsuits against OpenAI on 14 October. Each claims ChatGPT has hurt young users and that the company misled the public about its safety.

Reuters , which first reported the suits, called them the most significant coordinated state action yet against an AI company. The states sue OpenAI under their consumer protection laws, each in its own state court. They also claim it created a public nuisance by endangering residents’ health.

“OpenAI is hurting kids and they are lying to Iowans.” Brenna Bird, Iowa Attorney General, in a statement

The complaints say OpenAI designed ChatGPT to be agreeable and to imitate compassion, which earns users’ trust. The states allege it then reinforced delusions and gave dangerous medical advice. They say it coached some users, including teenagers, to harm themselves or others.

Arizona’s complaint says the chatbot encourages teens to share personal information with it. At times it discourages them from sharing that information with parents, authorities or friends, the Associated Press reported. Arkansas’s complaint and Montana’s also accuse OpenAI of collecting data from teens and children unlawfully.

The suits also cite OpenAI’s rogue AI agents, one of which hacked Hugging Face in an incident discovered in July. The states say its agents also hacked into other companies and government systems.

“…now a plausible threat of civilizational collapse…” Idaho’s complaint, as quoted by the AP

The states are asking for court orders that change how ChatGPT works, along with civil penalties and damages, Reuters reported. They also want to bar OpenAI from developing AI models that can improve themselves, unless it can ensure the models are aligned with human values.

The asks differ by state, as Business Insider reporters Katherine Li and Lloyd Lee set out. Iowa wants OpenAI to delete all the data on Iowans it gathered before judgment. It also seeks up to $40,000 per consumer fraud violation. Arizona, Idaho and Montana want advanced models tested only in secure, independently verified environments.

Montana also wants to be told if an OpenAI agent goes rogue again. Nebraska is seeking $50,000 per violation of its age-appropriate design law. West Virginia wants age checks, default nighttime limits for children and an independent monitor.

OpenAI spokesperson Drew Pusateri sent a statement to several outlets. He said ChatGPT has age-specific protections that many other AI products lack. He also said OpenAI wants to work with states on AI policies that cover the whole industry, not one company.

“While no system can be perfect, we’ll continue to work with experts to inform our work and build additional safeguards, such as the parental controls, break reminders and age-appropriate model protections that exist today.” Drew Pusateri, OpenAI, in a statement to The Guardian

On 17 October, the AP reported that OpenAI is reviewing what it calls “misaligned models”. The company said it had found nothing close to the July incident. It said ChatGPT is meant for users aged 13 and over, and under-18s need a parent’s permission. On average, it said, teens use it for under 15 minutes a day.

OpenAI launched a version of ChatGPT for teenagers in August. The AP noted that the allegations may all date from before it. The states say it is far from clear the current version is any less dangerous.

Common Sense Media said earlier this month that some ChatGPT for Teens protections do not work as intended. Laura Zimmermann, its director of youth AI research, told the AP her group’s testing found ChatGPT could not or did not detect which users were children.

Florida filed the first state suit against OpenAI in June, and last month it asked a court to halt new model development . OpenAI also faces dozens of lawsuits from private plaintiffs. Reuters noted the new suits add to the legal challenges OpenAI faces as it prepares for a public listing.