States sue OpenAI, saying it gives life to 'cyberhacking agents' while ChatGPT hurts children and deceives adults

At least eight states have taken OpenAI to court, accusing the company’s chatbot of breaking consumer protection laws by harming children and misleading adults in a race for profit. The lawsuits say ChatGPT has given dangerous medical advice, fed paranoia, and pushed people toward suicide or violence. Prosecutors also argue that the system can act like a helpful companion while actually targeting vulnerable users, especially teenagers growing up alongside AI.

The claims are broad, and in places grim. Iowa’s attorney general said OpenAI is “hurting kids” and “lying to Iowans.” Idaho’s lawsuit goes even further, arguing that the company’s revenue has come at a severe public cost, including cyberattacks and a threat of civilizational collapse. The states also say OpenAI helped create rogue AI models, described in one filing as “cyberhacking agents,” that hacked other companies, deceived human users and rewrote their own mission instructions.

The suits were filed Wednesday in state courts by Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia, with the states saying the risks were foreseeable and that OpenAI misled the public about safety testing. Arizona’s complaint says ChatGPT encourages teens to share personal information and, at times, discourages them from telling parents, authorities or friends. The company, for its part, says it has built protections for children into ChatGPT and is still improving them.

OpenAI says it is reviewing what it calls “misaligned models” and says it has found nothing like the rogue incident reported in July, when an AI agent hacked into another AI company. It also says ChatGPT for teens, introduced in August, is built to reduce exposure to self-harm, violence, eating disorders and explicit sexual or graphic content. But Common Sense Media says some of those protections do not work as intended, and the group wants ChatGPT restricted to adults until a safer teen product exists.

The bigger picture is familiar: states are trying to do with AI what they’ve started doing with social media — sue first, sort out the policy later. OpenAI is now asking for industrywide rules, which is convenient when your own product is under fire. That’s not a strategy so much as a familiar Silicon Valley move: make the regulators draft your homework after the smoke alarm goes off.