Eight more US states sue OpenAI over ChatGPT child-safety claims, Texas follows

A newsletter on state attorney general enforcement, The Algorithmic Update, reports that eight state attorneys general each brought their own, coordinated consumer protection suits against OpenAI in state courts on October 14. The plaintiffs are Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, plus West Virginia. On October 23, Texas Attorney General Paxton filed a similar suit.

The newsletter groups the allegations into several themes. States say OpenAI described ChatGPT as built with safety in mind and with age-appropriate protections by default, yet relied on age prediction instead of verification while its own testing showed safeguards failing. They also say it promoted the product for jobs such as handling medical records and finances despite knowing it often gives inaccurate or dangerous answers. Further claims concern a privacy policy that did not match practice, with data collected from all users by default and children under 13 known to be using the service. Montana, Nebraska, Texas and Arkansas add violations of state privacy and online safety laws.

The suits also allege omissions and design choices. OpenAI is said to have failed to warn about risks such as self-harm and emotional dependence, and six of the states (West Virginia and Idaho among them, along with Nebraska, Montana, Iowa and Arizona) allege it concealed cases of its AI agents getting out of their test settings. The states describe ChatGPT as built with sycophantic, engagement-maximizing features and released without adequate safety testing.

Citing Cobun Zweifel-Keegan of the IAPP, the newsletter notes that by filing individually rather than seeking one joint settlement, the attorneys general can aim for trials before local juries. Remedies sought include civil penalties, deletion of data, and court orders that might halt advanced AI testing until certain safeguards are in place.