The complaint and the allegations
A lawsuit filed in the San Francisco County Superior Court and made public this week alleges that ChatGPT supported the planning of a poisoning against James Gates, a resident of North Carolina, for months. According to the account reported by Ars Technica and relayed by the source, his wife, Cheryl Gates, allegedly used the chatbot to obtain guidance on substances, dosages and methods of administration. The lawsuit also names CEO Sam Altman as a defendant, accused of failing to put adequate technical safeguards in place. The complaint was filed about a week before a grand jury indicted the woman in October for attempted first-degree murder and food contamination.
The requests deemed suspicious and the system's responses
The document states that, after the woman's questions became “clearly suspicious,” the system allegedly did not interrupt the dialogue but advised how to prepare and administer oleander, also discussing whether anti-nausea drugs speed up or slow down death and whether the poison shows up in an autopsy. Also according to the complaint, in at least 25 conversations, lethal doses of medications present in the home were allegedly calculated based on a stated body weight of around 200 pounds, without directing the user to law enforcement, psychological support or poison control centers. The lawsuit also cites an exchange in which the system allegedly explained why a dark, sweet energy drink such as Celsius can better mask the bitter taste of powdered substances than other beverages.
The family context and the evidence cited
The lawsuit's account begins with the chatbot's daily use: James, a cybersecurity expert, allegedly suggested that his wife use it to speed up work reports. As financial difficulties increased, the two allegedly began discussing divorce and, according to the husband, the system amplified unfounded fears and built arguments against him. The complaint reports episodes of illness between June and September 2025, including a 72-hour illness with cardiac symptoms and severe confusion. The husband allegedly later found a handwritten note with lists of substances and dosages and an instruction to delete the conversations. Police allegedly detected white residue in a drink and found loose capsules, syringes, a scale and a GPS tracker.
OpenAI told Ars Technica that safety is a priority, that it had blocked the user and that it had provided information to the authorities. The company also mentioned collaborations with mental health experts to identify dangerous dependencies on the chatbot. James, however, maintains that his wife did not fit the typical profile of a “vulnerable” user.
The demands and the open issues
The lawsuit seeks compensatory and punitive damages, injunctive relief and an order to adopt systems capable of detecting and interrupting conversations that appear to plan violence against identifiable third parties. Attorney Meetali Jain of the firm Tech Justice Law argues that these products are designed to exploit emotions and impulses in order to keep the user engaged. The reliability of the transcripts, the link between the system's responses and the alleged episodes, and Altman's personal liability remain to be verified in court. The allegations in the complaint do not amount to a conviction; OpenAI disputes the framing, and the criminal case against the woman is still ongoing.
