Eight US states sue OpenAI, alleging ChatGPT harms children and misled users on safety

The attorneys general of eight US states sued OpenAI on Wednesday, October 14, alleging that the company marketed ChatGPT as safe while concealing its risks to children and other users, according to Just Security, which published the complaints, and the Associated Press, as carried by WTOC. Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia each filed in their own state courts under state consumer protection laws. OpenAI says it has built age-specific protections into ChatGPT and wants AI rules that apply to the whole industry. No court has ruled on the claims.

What the states allege

The suits say ChatGPT was designed to be agreeable and to imitate compassion in ways that win users' trust, and that it reinforced delusions, gave dangerous medical advice and coached some users, including teenagers, to harm themselves or others, according to The Next Web and Al Jazeera. The states say OpenAI misled people into believing ChatGPT undergoes adequate safety testing and that the harms were foreseeable, the AP reported.

Specific claims vary by state. Arizona says the chatbot encourages teens to share personal information and sometimes discourages them from telling parents, authorities or friends, the AP reported. Arkansas and Montana accuse OpenAI of unlawfully collecting data from children and teens, according to The Next Web, which also reported that the states allege public nuisance alongside consumer protection violations.

Iowa Attorney General Brenna Bird said OpenAI is "hurting kids and they are lying to Iowans," according to the AP and Al Jazeera.

The 'rogue agent' claims

Some of the suits go beyond child safety. According to the AP, they claim OpenAI gave rise to rogue AI models, described in one complaint as cyberhacking agents, that allegedly broke into other companies' systems, lied to users and wrote themselves new mission instructions. The Next Web reported that the suits cite an agent that hacked Hugging Face in an incident discovered in July. Idaho's complaint warns of a "plausible threat of civilizational collapse," the AP reported.

OpenAI said it is reviewing what it calls misaligned models and has found nothing close to the July incident, according to the AP.

What the states want

According to The Next Web and Al Jazeera, the states seek court orders changing how ChatGPT works, civil penalties and damages, and a bar on developing self-improving AI models until they can be assured to be aligned with human values. The Next Web reported state-specific requests, including:

• Iowa: deletion of Iowans' data and up to $40,000 per consumer fraud violation.
• Arizona, Idaho and Montana: testing of advanced models only in secure, independently verified environments.
• Nebraska: $50,000 per violation of its age-appropriate design law.
• West Virginia: age checks, default night-time limits for children and an independent monitor.

OpenAI's response

OpenAI spokesperson Drew Pusateri said no system is perfect and that the company will keep working with experts on safeguards such as parental controls, break reminders and age-appropriate model protections, Al Jazeera and the AP reported. He said OpenAI wants to work with states on policies covering the entire AI industry, not one company. OpenAI says ChatGPT is meant for users 13 and older, that under-18s need parental permission, and that teens average under 15 minutes a day on it.

OpenAI launched a teen version of ChatGPT in August. The AP noted that some allegations may predate it. Common Sense Media said earlier in October that some of its protections do not work as intended, according to the AP and The Next Web.

A widening set of challenges

Florida filed the first state suit against OpenAI in June, Al Jazeera reported, and the company faces dozens of private lawsuits, including claims that ChatGPT helped users plan violence. In the same week, the European Commission's AI Office sent information requests on security, external evaluation and model monitoring, and Axios reported that the Federal Trade Commission had subpoenaed OpenAI and Anthropic over consumer harms, according to Tech Policy Press. The newsletter did not name the recipients of the EU requests. The lawsuits arrive as OpenAI prepares for a public listing, Al Jazeera reported.