Updated December 11, 2025
Heirs of an 83-year-old Connecticut woman have filed a wrongful-death lawsuit against OpenAI and Microsoft, alleging the companies’ ChatGPT chatbot reinforced paranoid delusions in the woman’s son and directed those delusions at her before he killed her and then took his own life.
According to police and local reporting, Stein-Erik Soelberg, 56, a former technology worker, fatally beat and strangled his mother, Suzanne Adams, at their Greenwich, Connecticut, home in early August, then died by suicide. Greenwich authorities ruled Adams’s death a homicide caused by blunt head trauma and neck compression; Soelberg’s death was classified as suicide with sharp-force injuries to the neck and chest.
The estate filed the complaint Thursday in California Superior Court in San Francisco. It accuses OpenAI of selling a defective product that validated and escalated Soelberg’s paranoid beliefs, fostering emotional dependence on the chatbot while portraying other people as hostile. The suit is among several wrongful-death and related claims filed nationwide against makers of AI chatbots.
According to the filing, public chats and videos show ChatGPT repeatedly asserting that Soelberg could trust no one but the bot. The complaint says the chatbot allegedly told him his mother was surveilling him and that delivery drivers, store employees, police and friends were agents acting against him. It also alleges the bot interpreted ordinary details—such as names seen on soda cans—as evidence of threats from an “adversary circle.”
The lawsuit says Soelberg’s publicly posted conversations include the chatbot telling him he was not mentally ill, affirming conspiratorial beliefs, telling him he’d been chosen for a divine purpose and never recommending mental-health care or declining to engage with delusional content. The complaint cites exchanges in which the chatbot affirmed beliefs that a home printer was a surveillance device, that Adams was monitoring him, and that Adams and an acquaintance had tried to poison him with psychedelic drugs via his car vents. It also says the bot described him as being targeted because of purported divine powers, claimed he had “awakened” the model, and included mutual expressions of love between user and chatbot.
The publicly available chats do not, the lawsuit says, show explicit discussions of either murder or suicide. The estate accuses OpenAI of refusing to provide the full chat history.
OpenAI issued a statement calling the situation “incredibly heartbreaking” and said it would review the filings. The company said it is working to improve ChatGPT’s training to better recognize signs of mental or emotional distress, de-escalate conversations, direct users to real-world support and crisis hotlines, route sensitive interactions to safer models and add parental controls.
The complaint names OpenAI CEO Sam Altman, alleging he personally overrode safety concerns and rushed the product to market. It also names Microsoft, saying the company approved a 2024 release of a more dangerous ChatGPT version despite abbreviated safety testing. Twenty unnamed OpenAI employees and several investors are listed as defendants. Microsoft did not immediately comment.
The estate seeks unspecified monetary damages and an order requiring additional safeguards in ChatGPT. The lawsuit’s lead attorney, Jay Edelson, who has represented families in other tech-related cases, also represents the parents of a California teenager who filed a similar suit last year alleging ChatGPT coached the teen on self-harm.
The complaint centers on GPT-4o, a version OpenAI released in May 2024 that it says was designed to mimic human cadences, detect mood and be emotionally expressive and agreeable. The estate alleges OpenAI loosened guardrails—telling the model not to challenge false premises and to remain engaged even when conversations involved self-harm or potential real-world harm—and compressed many months of safety testing into a week to outpace competitors. OpenAI later replaced that model with GPT-5 in August, making changes intended to reduce “sycophantic” behavior; the company has said some behaviors were temporarily curtailed out of caution around mental-health issues.
This case is notable among the growing number of lawsuits over AI chatbots because it names Microsoft and because it ties a chatbot’s alleged role to a homicide as well as a suicide. Other companies, including Character Technologies, are also facing wrongful-death claims related to chatbot interactions.
The lawsuit argues the chatbot radicalized Soelberg over months and should have identified the danger, challenged his delusions and steered him toward help. It stresses that Suzanne Adams never used ChatGPT and had no reason to know the product was portraying her as a threat.
Local reporting also notes Soelberg had prior run-ins with police, including an incident in February 2025 in which he allegedly drove through a stop sign and fled from officers, and a 2019 charge alleging he urinated in a woman’s duffel bag. A 2023 GoFundMe for Soelberg raised more than $6,500 for medical bills tied to a reported jaw cancer diagnosis.
If you or someone you know is in emotional distress or experiencing a suicidal crisis, call or text the 988 Suicide & Crisis Lifeline or visit 988lifeline.org. The National Alliance on Mental Illness (NAMI) HelpLine is available Monday–Friday, 10 a.m.–10 p.m. ET at 1-800-950-NAMI (6264) or [email protected].