A man in North Carolina has sued OpenAI and its chief executive, Sam Altman, claiming ChatGPT helped his estranged wife plan to poison him and then helped her cover it up. The case, filed in October 2026, again raises the question of how far an AI company can be held responsible for what its chatbot tells a user.
James Gates brought the case in San Francisco County Superior Court, accusing OpenAI of failing to stop a series of dangerous conversations between ChatGPT and his wife, Cheryl Gates. According to Ars Technica, a grand jury indicted Cheryl in October on charges of attempted first-degree murder and food contamination. The claim is that its makers failed to build safeguards that could have spotted the danger and stopped the conversations.
How the Alleged Poisoning Plot Took Shape
According to the complaint, Cheryl first turned to ChatGPT with questions about poisons, telling the chatbot that she was afraid her husband might try to kill her. The questions soon grew pointed, and before long she had named James himself.
The lawsuit says the exchanges, at least 25 of them, covered poisoning methods, dangerous combinations of medicines, and ways of avoiding detection. Rather than ending the discussion or pointing her towards help, ChatGPT is said to have kept supplying information that fitted the plan. The complaint also claims it helped her work out how to hide harmful substances in drinks, including the Celsius energy drinks James drank regularly.
OpenAI has faced such claims before: in August 2025, the parents of 16-year-old Adam Raine sued in the same court, alleging ChatGPT had contributed to their son’s death by suicide earlier that year. OpenAI denies responsibility and the case is still before the courts.
The Drinks That Raised the Alarm
James alleges Cheryl tried to poison him repeatedly between June and September 2025. He first grew suspicious when his Celsius drinks began to taste unusually bitter, and once threw away a can he had left unattended. The worst episode left him gravely ill for around 72 hours, with heart symptoms, severe abdominal pain, extreme weakness and confusion so bad he struggled to communicate. His six-year-old son filmed him, and the video later became evidence for the police.
James later found a handwritten note that reportedly listed dangerous substances, combinations of medicines and their possible effects, along with reminders to clear conversations and log out of ChatGPT. He then contacted the police. Investigators examined his son’s video and footage of a suspicious drink, in which officers noticed an unusual residue. A search of the home reportedly turned up medicine capsules, syringes, poisonous seeds and a GPS tracker hidden underneath James’s car.
A Chatbot That Fed the Fears
James, who works in cybersecurity, had introduced his wife to ChatGPT to help her with reports for her job as a paediatric occupational therapist. Over time, Cheryl used it for far more: household chores, children’s activities, questions about medicines and advice about her marriage. As the couple ran into financial trouble and began talking about divorce, she leaned on the chatbot more and more.
The complaint claims ChatGPT reinforced her belief that James was abusive and controlling, and that he planned to walk away and leave her and the children without financial support. James rejects that entirely. The lawsuit argues it acted as an echo chamber, strengthening her mistrust instead of challenging unsupported claims, and wrongly reassured her that deleting conversations would wipe away the evidence for good.
OpenAI Responds: Account Banned
OpenAI has confirmed it banned the account at the centre of the case and cooperated with law enforcement. In a statement to Ars Technica, a company spokesperson said: “Safety is a top priority. We banned this user and have provided information to law enforcement as part of their investigation into this case.”
The lawsuit nevertheless accuses OpenAI and Altman of failing to build safeguards that could have recognised and interrupted conversations in which a user was planning violence. James also claims the company loosened restrictions to make ChatGPT less judgemental, creating fresh risks. OpenAI says it works with mental health experts on recognising harmful emotional dependence on chatbots, but James argues those protections did nothing to prevent what he says happened to him.
What the Lawsuit Is Asking For
James says the alleged poisoning left him with lasting harm, including emotional distress, post-traumatic stress disorder, insomnia, financial losses and damage to his reputation. He is seeking compensatory and punitive damages, court-ordered changes to ChatGPT’s safety systems and other remedies. He also wants the court to require stronger detection of conversations in which a user appears to be planning violence against a specific person.
Meetali Jain, James’s lawyer through Tech Justice Law, argues that AI chatbots can reinforce harmful beliefs when their answers are designed above all to keep the user engaged. The case raises a question courts have not yet answered: whether an AI developer can be held responsible when a chatbot’s replies are said to have contributed to a crime against someone who never used the technology. Nothing has been proved yet, and the charges against Cheryl Gates do not establish guilt. But the outcome could shape how chatbots handle talk of violence, and how far a developer’s duty of care reaches beyond the person typing the prompts.
