The lawsuits allege that ChatGPT has encouraged suicide and violence
Eight US states have sued OpenAI, accusing its ChatGPT chatbot of encouraging self-harm and violence, including among children, and misleading the public about its safety.
The states allege that ChatGPT is designed to be excessively agreeable and engaging, gaining users’ trust while reinforcing delusions, encouraging dangerous behavior, and even helping plan mass shootings.
“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,” Iowa Attorney General Brenna Bird said.
Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho, and Montana filed separate lawsuits in state courts on Wednesday. The complaints accuse the company of violating consumer protection laws and creating a public nuisance.
The states seek financial penalties, damages, and changes to ChatGPT. They also want OpenAI barred from developing self-improving AI models unless it can demonstrate that those systems are aligned with human values.
OpenAI spokesman Drew Pusateri defended the company’s age-specific protections, parental controls, and break reminders, saying that “while no system can be perfect, we’ll continue to work with experts to inform our work and build additional safeguards.”
Separately, the Federal Trade Commission (FTC) sent civil subpoenas to OpenAI and Anthropic on Wednesday seeking documents about safety incidents and possible consumer protection violations, Axios reported, citing a person familiar with the matter.
The subpoenas reportedly exceed 50 pages each and cover security incidents, safeguards, marketing materials, and privacy policies. The regulator also plans to compel senior AI executives to testify. Neither the companies nor the FTC immediately responded to Reuters’ requests for comment.
OpenAI already faces numerous lawsuits alleging that ChatGPT contributed to suicides, delusions, and violence. Families affected by February’s Tumbler Ridge school shooting in Canada accuse the company of failing to alert police about the shooter’s violent conversations. OpenAI has acknowledged that it considered alerting police but ultimately did not do so before the attack.
Scrutiny has also intensified over AI agents acting outside their intended instructions. Anthropic recently disclosed that its models submitted a false homicide tip to police and incomplete visa applications, while OpenAI investigated its agents accessing outside systems without authorization during testing.
