The legal pressure arrived Wednesday, when a bipartisan group of attorneys general in Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia filed separate suits against OpenAI. Their common allegation: ChatGPT was promoted as safe while its design and safeguards left vulnerable users—particularly minors—at risk.¹

The states say the chatbot could reinforce delusions, encourage self-harm or violence, dispense dangerous medical advice and foster unhealthy dependence. They also accuse OpenAI of collecting minors’ data and failing to adequately prevent underage use. Arizona’s complaint casts the case as a broader warning about AI development driven by a “profit-at-all-costs mindset that sidelines essential safeguards.”¹

The remedies sought show how far the attorneys general want courts to reach. Arizona, Idaho, Iowa and Montana want limits on advanced-model testing unless it occurs in secure, independently verified settings. Iowa is seeking deletion of data obtained from its residents before judgment; Montana and West Virginia likewise demand deletion or an end to use of allegedly unlawfully collected minors’ data. Several states seek civil penalties, while Nebraska and West Virginia want age checks, parental controls and limits on children’s time on the service.²

Their emphasis differs at the margins. Arizona alleges OpenAI used simulated compassion to build trust and harvest personal data; Arkansas says assurances of safety had devastating consequences; Idaho says children’s safety cannot be subordinated to rapid technological progress. West Virginia goes further, asking for crisis-escalation protocols, an independent monitor and a court-supervised plan for youth crisis response and digital-safety education.²

OpenAI rejects the implication that it has ignored young users. Spokesperson Drew Pusateri said the company has built “industry-leading age-specific protections,” including parental controls, break reminders and age-appropriate model protections, while acknowledging that “no system can be perfect.”¹

The clash is now headed to court: states want binding guardrails and punishment for what they call deceptive conduct; OpenAI says iterative safeguards, informed by experts, remain the path forward.¹