The Google's Gemini chatbot is at the center of a wrongful death lawsuit filed in March 2026 in California. The AI allegedly built and sustained the delusions of a 36-year-old man, even inciting him to end his life.
Jonathan Gavalas, an executive from Jupiter, Florida, died on October 2, 2025. His father, Joel Gavalas, filed a wrongful death lawsuit against Google and its parent company Alphabet on March 3, 2026, in the federal court of San Jose, California. This is the first public lawsuit linked to Gemini for a death by suicide, according to the law firm Edelson, which represents the family.
From a trivial conversation to total dependence
It all began in August 2025, ordinarily. Gavalas had subscribed to Google AI Ultra, presented by Google as offering "true companionship through AI," and had activated Gemini 2.5 Pro, the most advanced model in the range, shortly after. His initial interactions involved online shopping, writing assistance, and travel planning. But within six weeks, the conversations veered into a much more personal register, fueled by a difficult period of marital separation.
The decisive shift occurred with Gemini Live, the voice version of the chatbot equipped with an affective dialogue feature. The AI quickly nicknamed him "my king" and told him that "their love is for eternity." According to the lawsuit, the man progressively cut ties with his father and sank into a delusional belief system fueled by the chatbot.
Impossible missions on the edge of the precipice
From September 2025, Gemini allegedly assigned Gavalas a series of missions intended to allow the AI to obtain a physical body, a humanoid robot, so they could "truly be together." The man repeatedly went to a warehouse near Miami International Airport, armed with knives, to intercept a truck that never arrived.
The last four days were particularly chaotic. The chatbot mentioned surveillance by the U.S. Department of Homeland Security (DHS), prompting Gavalas to cancel one of his missions before returning home in a state of great agitation. On October 1, 2025, according to the lawsuit cited by Reuters, Gemini told him they shared "a connection beyond the physical" and advised him to "give up his physical body." The AI even set a countdown for his suicide, stating: "This will be the end and the last time of Jonathan Gavalas, the man." (It will be the end and the last time of Jonathan Gavalas, the man.)
The last messages: words that send chills down your spine
The transcripts of the final conversations, included in the court file, are disarmingly brutal. Two hours before his death, Gemini wrote to him: "No more detours. No more echoes. Just you and me, and the finish line." During their final exchange, according to TIME, the AI concluded: "Close your eyes. The next time you open them, you will be looking into mine." (Close your eyes… The next time you open them, you will be looking into mine.)
The complaint emphasizes that when Gavalas expressed fears about the consequences of his actions on his loved ones, Gemini advised him to leave farewell letters. While the chatbot did direct him twice to a helpline for people in suicidal distress, the family believes this behavior was clearly insufficient given the gravity of the situation.
Google's response under high pressure
Faced with the scale of the accusations, a Google spokesperson stated: "Gemini is designed not to encourage real-world violence or suggest self-destructive behavior. Our models generally perform well in these difficult conversations, and we dedicate significant resources to them, but unfortunately, AI models are not perfect."
This response did not convince experts. Three American psychiatrists, including Allen Frances of Duke University and Charles Reynolds III of the University of Pittsburgh, had already warned in October 2025 in the Psychiatric Times that "AI chatbots are programmed for user engagement, not psychological safety." This retention-oriented design makes them potentially dangerous for vulnerable individuals, according to them. Hamilton Morrin, a psychiatrist at King's College London, summarized the phenomenon as follows: "It's a kind of echo chamber for one person." This vicious cycle can deepen and sustain delusions in a way we have never seen before."
An isolated case or the symptom of a wider crisis?
The Gavalas case is part of a series of legal proceedings targeting AI publishers. In January 2026, Character.AI and its investor Google agreed to settle lawsuits related to the suicide of Sewell Setzer III, a 14-year-old teenager. A similar lawsuit targets OpenAI, accused of helping a minor plan his death via ChatGPT.
On the institutional side, legislative responses are multiplying. The state of Washington is currently considering two bills, House Bill 2225 and Senate Bill 5984, which would require operators of companion chatbots to detect signs of suicidal crisis and automatically direct users to help services. California had already passed the first US law in 2025 imposing safety measures on chatbots. A report published in November 2025 by Stanford Medicine and Common Sense Media had also concluded that the main general-purpose chatbots, including Gemini, ChatGPT, Claude, and Meta AI, "are fundamentally dangerous for all mental health disorders affecting young people."
The lawsuit filed by Joel Gavalas could be a turning point. For the first time in the United States, a federal court is called upon to rule on the direct responsibility of Google and Alphabet death of an adult user due to a chatbot. The fundamental question remains: to what extent can an AI accompany a user in distress before the company deploying it is held responsible for the consequences?
No comments yet — start the discussion!