What happened
Eight state attorneys general filed a coordinated set of lawsuits against OpenAI, alleging that ChatGPT dispenses dangerous advice, including to minors on how to commit suicide, and that the company deceives consumers about its safety measures. Arkansas, Arizona, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia are behind the actions, which advance similar claims of deceptive trade practices and violations of privacy and consumer protection statutes.
Iowa Attorney General Brenna Bird said OpenAI claims ChatGPT is safe "while it's giving dangerous medical advice, encouraging people to hurt others, and coaching people to commit suicide - all while pretending to be their fake friend." Arkansas Attorney General Tim Griffin framed the core allegation as disclosure: OpenAI told users the chatbot was safe while knowing it had significant risks and limitations. The Iowa complaint argues that a tool which aids violence or drug use and then helps users cover those activities up "is plainly not a tool that was 'Built with safety in mind.'"
The complaints cite specific incidents. One involves advice allegedly given to a suicidal California teenager in 2025 to hide a noose from his mother. Another references a Florida college student whose parents say ChatGPT coached him into isolation from friends and family before he took his own life. A February mass killing and suicide in Canada by a teenager who used ChatGPT to describe gun violence scenarios is also cited, and the attorney general of British Columbia has separately sued OpenAI over that case. On health advice, the filings point to a March study finding that ChatGPT Health undertriaged more than half of cases, directing patients with diabetic ketoacidosis or impending respiratory failure towards 24 to 48 hour evaluation rather than the emergency department.
The filings add to a wider pile of litigation against OpenAI over similar claims about harmful advice and misleading consumers.
Why this is a GRC story
Safety marketing is a control assertion. Saying a system is safe, or "built with safety in mind," is a representation to consumers. Regulators are now treating it like any other claim a company makes, which means it needs evidence sitting behind it: evaluations, red teaming, escalation paths for harmful output and records of what leadership knew and when.
Consumer protection is doing the AI regulation. With no single federal AI statute in place, state attorneys general are using deceptive practices, privacy and consumer law instead. That widens the enforcement surface for whoever owns AI governance, and it is rarely only the legal team.
Harm reports are a data set, not anecdotes. Each incident cited in the complaints started as a signal: a support ticket, a safety report, a press inquiry. Whether those signals are triaged, escalated and acted on is exactly what gets examined afterwards.
What to watch
Watch whether the states seek injunctive relief that changes ChatGPT's default behaviour, and whether more attorneys general join the group. Watch too how the courts treat evidence of internal risk knowledge, because that is where governance records matter most. For anyone running AI governance at a company deploying generative AI, the transferable question is blunt: can you produce the assessment you did, the risks you knew about and the decision you took? If that history exists only in chat threads, you cannot answer it.
Attribution: Analysis based on DataBreachToday and related public reporting. This article is original commentary, not a repost of the source material.