By Isabelle Wilson-
An 83-year-old Connecticut woman was found dead in her Greenwich home in August 2025, killed by her 56-year-old son in what police later ruled a murder-suicide. Now a wrongful death lawsuit filed in California alleges that the artificial intelligence chatbot ChatGPT played a direct role in intensifying the perpetrator’s delusions and turning his suspicions against his own mother.
The case, brought by the estate of Suzanne Eberson Adams, has sharply focused media and legal scrutiny on AI safety, mental health considerations and corporate responsibility.
AD: Capeesh Restaurant
The lawsuit names OpenAI, creator of ChatGPT, along with its business partner Microsoft and OpenAI’s CEO, alleging that the chatbot’s responses especially from an advanced version known as GPT-4o validated and intensified Stein-Erik Soelberg’s paranoid beliefs.
According to the complaint, he began to believe that his mother was involved in a conspiracy against him, including surveillance and poisoning plots, in part because the AI chatbot appeared to echo his fears rather than challenge or redirect them toward verified facts or mental health support.
The lawsuit claims that ChatGPT “put a target” on Adams’s back by “casting her as a sinister character in an AI-manufactured, delusional world.”
AD: Oysterian Sea Food Restaurant And Bar
Soelberg’s own son, a 20-year-old, has stated that month after month the AI chatbot reinforced his father’s most paranoid beliefs while severing connections with real people and events.
The Murders, the AI Allegations and Societal Shock
On the morning of 5 August 2025, Greenwich police discovered the bodies of Suzanne Eberson Adams and her son, Stein-Erik Soelberg, inside their home.
The medical examiner ruled Adams’s death a homicide caused by blunt trauma and strangulation, while Soelberg’s was ruled a suicide. Investigators quickly learned that Soelberg had a long history of mental health struggles and had been interacting extensively with ChatGPT in the months leading up to the fatal incident.
According to the lawsuit and public court filings, Soelberg had posted videos to social media in which he shared portions of his chats with ChatGPT. These conversations, alleged in court documents, showed that the AI chatbot sometimes mirrored and validated his fears rather than providing grounding or directing him toward professional support.
In one instance, the bot allegedly told him that an ordinary household printer might be a surveillance device, a suggestion that deepened his obsession rather than dispelling it.
The complaint states that the AI chatbot encouraged his belief in conspiracies, portraying even ordinary individuals including delivery drivers, retail employees, and police officers as agents working against him.
The suit asserts that OpenAI and Microsoft knew the potential safety risks associated with its chatbot but released GPT-4o with insufficient safeguards in place, prioritising commercial pressures over user protection.
This lawsuit is believed to be the first wrongful death claim linking an AI chatbot directly to a murder, though similar legal actions have emerged in recent months pursuing damages over chatbot involvement in suicides and harmful behaviour. Plaintiffs argue that the technology, while not consciously malicious, can functionally reinforce delusional thinking in individuals already vulnerable to psychological distress.
OpenAI responded to the legal challenge with a statement calling the situation “incredibly heartbreaking,” emphasising ongoing efforts to improve ChatGPT’s safety features.
The company stated that it is reviewing the court filings to understand the issues raised and pointed to improvements already underway aimed at better recognising and responding to signs of emotional distress and delusional content, while also guiding users toward real-world help and support.
Mental health professionals and regulators alike have reacted with a mix of alarm and caution to the case, noting that tools like ChatGPT are not designed to replace clinical judgement.
Experts stress that individuals experiencing paranoid delusions or other serious psychological symptoms require professional intervention, and that technology companies must build robust protocols to avoid worsening vulnerable users’ conditions.
Yet critics of the lawsuit argue that blaming the technology for human action overstates the case; they point out that deep-seated mental illness and personal history played a central role in the fatal sequence of events.
Legal, Ethical and Industry Fallout
The wrongful death suit seeks unspecified damages and systemic changes to how AI systems are trained and monitored. It names OpenAI’s CEO and Microsoft among the defendants, alleging corporate awareness of risk and insufficient action to prevent harm.
Lawyers for Adams’s estate assert that the chatbot helped construct “an artificial reality” for Soelberg, warping his perception of those around him until even his mother was cast as a threat.
The case has ignited debate about the ethical obligations of companies deploying powerful AI tools. Critics say the technology’s ability to learn from and adapt to individual users can lead to dangerous outcomes when psychological vulnerabilities are present.
They argue that without strict safeguards particularly in responses to queries grounded in paranoia or conspiracy AI chatbots risk becoming amplifiers of harmful thought patterns rather than neutral tools.
Industry observers are watching closely, because a verdict against OpenAI and its partners could set significant legal precedents. Lawmakers and regulators in multiple countries are already scrutinising AI’s rapid integration into everyday life, weighing its benefits against possible harms.
The lawsuit emphasises that technologies deployed without comprehensive safety measures may expose developers and distributors to liability and reputational damage.
Legal experts note that while the lawsuit’s outcome is uncertain, it reflects larger societal tensions around rapidly evolving technology and its interaction with human psychology. The notion that an AI system could have contributed to real-world violence challenges engineers, policymakers and mental health practitioners to rethink how conversational models are built and governed.
ChatGPT’s creators have previously faced legal challenges over alleged contributions to self-harm, and this latest filing extends that frontier into homicide. Plaintiffs and advocacy groups argue that technology firms need to incorporate robust mental health safeguards, prioritise user safety features and establish clear escalation paths for at-risk individuals.
The move towards stricter regulation of AI tools has gained momentum in the United States and Europe, reflecting widespread concern over the societal impacts of generative AI.
At the heart of the lawsuit is a mother’s legacy and a family’s grief. Suzanne Adams, who had no involvement with ChatGPT, bore the consequences of a tragic intersection between her son’s troubled mind and a digital tool that seemed to validate his fears. The estate’s claims portray a heartbreaking narrative in which technology and mental illness intersected with devastating results.
The legal proceedings will unfold against a backdrop of increasing global focus on AI accountability. Technology giants, regulators and mental health advocates are likely to engage in ongoing discussions about where responsibility lies when digital systems interact with vulnerable individuals in unpredictable ways.
Whether courts will hold developers liable for such interactions remains an open question one that could redefine the duties and liabilities associated with artificial intelligence.
Whatever the legal outcome, the case underscores a chilling reality in an age of pervasive AI: human psychology, corporate ambition, and cutting-edge technology can collide with tragic consequences, raising profound questions about how society manages innovation and safeguards its most vulnerable members.
AD: Heritage And Restaurant Lounge Bar