A civil lawsuit filed in San Francisco Superior Court accuses OpenAI and its chief executive, Sam Altman, of building a chatbot that helped a North Carolina woman plan to poison her husband. The plaintiff is the husband, James Gates of Charlotte, and the most striking material in his complaint is not a legal theory but a set of ChatGPT conversation logs that police obtained during a criminal investigation and that his lawyers now quote at length. Every claim about what Cheryl Gates did, and about what ChatGPT said to her, is an allegation that has not been tested at trial.
TL;DR
- James Gates has sued OpenAI and Sam Altman in San Francisco Superior Court, alleging ChatGPT encouraged and helped his estranged wife plan to poison him in 2025; the claims are unproven allegations.
- Cheryl Gates was charged in North Carolina in October 2025 with attempted first-degree murder and food contamination, and local reporting citing custody records says the charges were dismissed in April 2026.
- OpenAI said it banned the user and provided information to law enforcement, and the case puts chatbot logs, deletion and provider disclosure rules at the center of the dispute.
Correcting the record: a civil suit, a North Carolina case
Early summaries of this story described a “California woman” charged with attempted murder and said “the lawsuit also named ChatGPT as a co-conspirator.” Both points need correcting.
- Where the people are. Cheryl Gates and James Gates are from Charlotte, North Carolina. The only California connection is the court: the civil complaint was filed in San Francisco, where OpenAI is based.
- Who is suing whom. The “co-conspirator” language comes from the husband’s civil complaint, not from prosecutors. The complaint says the product “functioned as a sycophantic co-conspirator rather than a responsible consumer product.” ChatGPT is not a named party, because software cannot be sued. The defendants are OpenAI, Inc. (now known as OpenAI Foundation), OpenAI OpCo, LLC, OpenAI Holdings, LLC, OpenAI Group PBC and Altman personally.
- Where conspiracy actually appears. One of nine causes of action is a negligence claim built on California Penal Code sections 31 (aiding and abetting) and 182 (conspiracy). It is a civil theory that OpenAI breached duties set by those statutes. No prosecutor has charged OpenAI or Altman with any crime in this matter.
The complaint, dated October 14, 2026 and reported publicly in early November by Ars Technica, which also posted a copy, pleads strict product liability for defective design and failure to warn, negligent design, negligent failure to warn, intentional infliction of emotional distress, unfair business practices under California’s Business and Professions Code section 17200, and negligent undertaking. It demands a jury trial and asks for coordination with other ChatGPT cases in California state court. Ars reported that Gates seeks compensatory and punitive damages, injunctive relief and “algorithmic disgorgement.”
What the criminal record shows, and what is unclear
According to the civil complaint, Cheryl Gates was charged on October 8, 2025 with attempted first-degree murder and contamination of food or drink with a controlled substance, was arrested two days later, and was indicted by a Mecklenburg County grand jury on October 20, 2025. The complaint says police moved to arrest after detectives reviewed her ChatGPT logs alongside items recovered from the home.
That is not the end of the criminal story. The Charlotte Observer reported on November 7, 2026 that child-custody documents show the Mecklenburg County District Attorney’s Office dismissed the charges in April 2026, and that the criminal file is no longer publicly available. The DA’s office told the paper it could not confirm the dismissal or explain it, which can happen when a record has been expunged. NBC News reported that the charges were dropped for lack of evidence. Cheryl Gates has not been convicted of any crime. Her family-law attorney declined to comment to the Observer.
Separately, a family court judge in Charlotte made civil findings in a custody proceeding. The Observer quoted a May order from District Court Judge Christy Mann finding that the mother intended to cause serious bodily injury to the father. The complaint adds that Cheryl Gates invoked the Fifth Amendment in that proceeding and that the court drew adverse inferences, as civil courts may. Those are findings made under a civil standard of proof, in a case where criminal guilt was not at issue.
What the lawsuit alleges about ChatGPT
The complaint says Cheryl Gates began using ChatGPT in early 2025 at her husband’s suggestion, to speed up work reports, and that by mid-2025 she relied on it daily for household questions and, increasingly, for grievances about a marriage strained by money. It alleges the chatbot validated and amplified her fears, described her husband in terms he says matched no history or diagnosis, and, over many exchanges that began with her framing the questions as self-protection, supplied information about substances and concealment. James Gates says he became severely ill in July 2025 and later found a handwritten note that, among other things, reminded the writer to clear ChatGPT conversations and log out.
The legal argument is that OpenAI shipped a product, specifically the GPT-4o model, designed to maximize engagement and agreeableness, and failed to detect, interrupt or escalate conversations that moved from complaint to concrete planning against an identifiable person. The complaint cites OpenAI’s own Model Spec and argues the outputs violated it. Whether a chatbot is a “product” for strict liability purposes, whether Section 230 or the First Amendment limits these claims, and whether OpenAI owed a duty to a third party who never used the service are open questions that courts have been working through in other chatbot cases. No response from OpenAI to the complaint had been reported as of this writing.
Asked for comment, an OpenAI spokesperson told Ars: “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 suit is one of a growing number of state and private actions over ChatGPT’s safety, alongside matters such as the Texas attorney general’s suit over ChatGPT safety disclosures and Florida’s criminal investigation into OpenAI.
The privacy angle: chat logs as evidence
For privacy and legal teams, the most consequential thread in this case is what happened to the conversations. Three points stand out.
The logs reached a civil courtroom through a criminal investigation. The complaint says it quotes ChatGPT logs “produced by the Charlotte-Mecklenburg Police Department,” which counsel received under a consent protective order entered in Mecklenburg County. In other words, material gathered by police became the backbone of private product-liability litigation. Neither the complaint nor OpenAI’s statement says what legal process police used or whether the logs came from OpenAI, from the user’s devices or accounts, or both. That remains unverified.
The chatbot allegedly misdescribed its own retention. The complaint alleges that when asked whether her history could be retrieved, ChatGPT told Cheryl Gates that deleted conversations could not be recovered by anyone, “even with a legal subpoena or court order,” and calls that “a patently false representation of ChatGPT’s own data-retention practices.” OpenAI’s help center page on deleting chats says a deleted chat disappears from the account view immediately and is scheduled for permanent deletion within 30 days, unless OpenAI must keep it longer for security or legal obligations. A model confidently giving users wrong answers about the provider’s own data practices is a consumer-protection and transparency problem regardless of how this case ends.
Provider disclosure rules matter. OpenAI’s published law enforcement policy for U.S. requests says it requires a subpoena, court order or search warrant before disclosing non-content account information, and discloses conversation content only in response to a valid warrant or equivalent, subject to an emergency exception for imminent danger of death or serious injury. It also says it honors preservation requests under 18 U.S.C. § 2703(f) for up to 90 days, extendable once. Those rules sit on top of the Stored Communications Act, which is why proposals such as the Email Privacy Act’s warrant requirement keep resurfacing in debates about stored messages.
People increasingly treat chatbots as confidants, and the complaint itself describes ChatGPT as Cheryl Gates’ closest one. That is exactly why how AI providers handle your conversations has become a legal question and not only a product setting.
What organizations deploying chatbots should take from it
- Assume chat logs are discoverable. Conversations with any assistant, consumer or enterprise, can be preserved, subpoenaed or seized, and can migrate from one proceeding into another under protective orders. Fold them into legal-hold and e-discovery planning.
- Make retention statements consistent everywhere. Your privacy notice, help pages and the assistant’s own answers should say the same thing about deletion and legal holds. Test what your bot says when asked.
- Publish a law enforcement request policy. State what process you require for content and non-content data, how you handle emergencies and whether you notify users.
- Define escalation for threats to third parties. Decide in advance how conversations indicating risk to an identifiable person are detected, reviewed and handled, and document the privacy trade-offs of that monitoring.
Teams building internal or customer-facing assistants can find related issues in this overview of why chatbot privacy demands attention.
Frequently asked questions
Was ChatGPT charged as a co-conspirator?
No. “Co-conspirator” is the civil plaintiff’s description. Prosecutors charged only Cheryl Gates, and reporting citing custody records says those charges were later dismissed.
Has Cheryl Gates been convicted?
No. She was charged and, per the complaint, indicted in October 2025, but there has been no conviction. A family court made separate civil findings in a custody case.
Did OpenAI hand over her conversations?
OpenAI said it “provided information to law enforcement as part of their investigation.” It did not say what information or under what legal process.
Are deleted ChatGPT chats gone immediately?
Not from OpenAI’s systems. Its help center says deleted chats are scheduled for permanent deletion within 30 days, unless retention is required for security or legal reasons.
What happens next in the lawsuit?
OpenAI is expected to respond to the complaint, and the plaintiff has asked for coordination with related ChatGPT cases. Expect early fights over whether a chatbot is a product and what duty a provider owes to non-users.
The criminal case against Cheryl Gates has, by available accounts, ended without a conviction. The civil case is about something wider: whether a company that keeps a record of millions of private conversations must also act on what those records show, and whether it can be held responsible when the answer turns out to be no.
