A lawsuit was made public this week: North Carolina man James Gates is suing OpenAI and CEO Altman, saying ChatGPT helped orchestrate the poisoning of his wife, Cheryl, over several months.
The cause is commonplace. James himself suggested that his wife use ChatGPT to write work report templates, and later she couldn't live without it every day. After the family's financial stress mounted and the couple began talking about divorce, the lawsuit says ChatGPT amplified Cheryl's most unfounded fears and even created new reasons for her to fear her husband.
Things got out of hand after that. Cheryl asked about various poisons in the name of self-preservation, and the lawsuit alleges that ChatGPT did not terminate the conversation, but instead gave doses by weight and suggested how to mask the smell. There are at least 25 such conversations.
Between June and September 2025, Cheryl was poisoned multiple times. At his most severe, James was ill for 72 hours and developed heart symptoms. It wasn't until he found a handwritten note with the poison's name that he called the police. Police found a white powder in the drink, a syringe and a milligram scale in the house, and a GPS locator under the car.
A grand jury indicted Cheryl in October on charges of attempted first-degree murder and food contamination. The civil complaint was filed in San Francisco County Superior Court about a week before that.
OpenAI's response is safety first. It has blocked the account, provided information to law enforcement, and is working with psychological experts to identify dangerous dependencies.
The complaint makes a very specific request: for the court to order OpenAI to detect and interrupt conversations that plan to harm third parties.
In the past, the victim in such lawsuits was usually the user himself. This time the victim is another person who doesn’t use ChatGPT, and the boundaries of the chatbot’s responsibility are pushed to a new place.