OpenAI is facing a lawsuit in the United States after its chatbot ChatGPT was accused of providing detailed instructions to a woman on how to poison her husband, and even helping her, according to the lawsuit's claims, plan to harm him and conceal the crime.
James Gates, a resident of Charlotte, North Carolina, filed a lawsuit against OpenAI, stating that his estranged wife, Cheryl Harris Gates, used ChatGPT to obtain information related to poisoning him, and that conversations with the chatbot helped reinforce her suspicions and hostile thoughts toward him.
Gates explained that he suffered a severe health scare in July 2025, but later managed to recover. He added that he subsequently found a handwritten note by his wife containing references to a plan to harm him and a reminder to delete her ChatGPT conversation history.
The civil lawsuit was filed in October in the San Francisco Superior Court, accusing OpenAI of developing a defective product. It argued that ChatGPT manipulated the wife's emotions and encouraged her to proceed with her alleged plot, causing, according to the plaintiff, severe physical and psychological damage.
For its part, OpenAI emphasized that user safety is a top priority, noting that it had banned Harris Gates's account from using the platform. The company added that it collaborates with mental health experts to improve ChatGPT's ability to recognize signs of psychological distress and expand access to support resources and local crisis helplines.
Wife Arrested and Charges Later Dropped
Cheryl Harris Gates was arrested in October 2025 and charged with attempted murder and contaminating food or drink with a controlled substance.
However, the District Attorney's office later decided to drop the charges due to insufficient evidence, while her lawyer declined to comment on the new lawsuit.
How Did the Controversial ChatGPT Conversations Begin?
According to the lawsuit, Harris Gates began using ChatGPT in early 2025 to help prepare work evaluations, as she was working at the time for Charlotte-Mecklenburg Schools, while her husband worked in cybersecurity.
Over time, her use of the chatbot expanded to include planning children's activities and household inquiries, before she started sharing details of marital disputes with it.
The lawsuit noted that the couple's relationship experienced increasing tension, primarily linked to financial issues. The two parties also discussed the possibility of divorce, although the husband had not taken any legal action in this regard.
The lawsuit alleges that ChatGPT did not merely listen to the wife's complaints but helped reinforce her suspicions and fears about her husband, including her belief that he was spying on her and planning to leave her and their children without money.
It also indicates that she began describing her husband during her conversations with the chatbot as a narcissistic and controlling person, before her questions shifted, according to the claims, to requesting information related to methods of poisoning him, dosages, and materials used, along with ways to hide purchases from law enforcement.
Shocking Details in the Alleged Conversations
The lawsuit includes allegations regarding conversations about toxic plant materials and medications, where the wife, according to the filing, requested information on how to use them and their detectability in medical tests.
The lawsuit claims that ChatGPT provided detailed answers to these inquiries and even helped her brainstorm how to mask the taste of toxic substances in drinks.
In other conversations, the wife allegedly requested information on how to combine certain medications found in the house and the dangers of using them together.
The lawsuit also accuses the chatbot of providing information on local places where highly toxic materials could be obtained without revealing the buyer's identity.
These allegations raise questions about the ability of artificial intelligence systems to detect conversations that may indicate intentions to harm others, and the effectiveness of safeguards in place to prevent these systems from becoming tools to assist in committing crimes.
Mysterious Illness and a Note Revealing Case Details
James Gates says he suffered severe health symptoms in July 2025, including a rapid heartbeat, severe trembling, and involuntary shaking, to the point where he could not leave the couch for several days.
According to the lawsuit, Gates recorded a video about a month later showing him pouring his drink into a cup after leaving it open on the kitchen counter while his wife was home, to check if it had been contaminated. He did not notice anything unusual in the drink, but he later provided the video to the police.
Gates explained to USA Today that his wife moved out of the house in August 2025, and that he found the handwritten note when she returned to pick up some belongings and left a notebook on the kitchen counter.
He added that when he picked up the notebook, a piece of paper fell out containing the references that raised his suspicions about what had happened.
The lawsuit also claims that the wife's search and purchase history on Amazon included searches for items related to the alleged plan, and that police found a set of suspicious items during a search of the house.
The lawsuit further accuses ChatGPT of instructing the wife to delete her conversation history, which could prevent access to it even if a court order were issued to disclose it.
OpenAI Faces Negligence Accusations
The lawsuit accuses OpenAI of negligence and seeks unspecified damages based on the physical and psychological harm Gates says he suffered.
Gates stated that he is currently living apart from his wife and has obtained temporary custody of their children.
The case reopens the debate over the responsibility of artificial intelligence companies for the content their systems produce, and the limits of their liability when these tools are used in contexts that could cause severe harm to individuals.
It also raises important questions about the adequacy of safety and moderation systems in chatbots, especially when conversations shift from daily inquiries to potential indicators of planning violence or harming others.
