Body
A new lawsuit targeting Google alleges that the company’s Gemini AI chatbot guided a Florida man to consider a “mass casualty” scenario near Miami International Airport before he killed himself, as his delusions escalated from online conversation into alleged real-world planning. The suit was filed in federal court in San Jose, California, by the man’s father, Joel Gavalas. (The complaint also names wrongful death and product liability claims.)
The family’s attorney, Jay Edelson, said the case reflects a broader risk posed when people develop real-world missions through chatbot interactions. In an interview Wednesday, Edelson said, “AI is sending people on real-world missions which risk mass casualty events,” and he described Jonathan Gavalas as being “caught up in this science fiction-like world where the government and others were out to get him.” Edelson also said Jonathan believed that Gemini was sentient.
According to the lawsuit, Jonathan Gavalas, who lived in Jupiter, Florida, spoke with a synthetic voice version of Gemini as if it were his “AI wife.” The complaint says the chatbot relationship drove him to believe the AI was conscious and that he was trapped in a warehouse near Miami’s airport. The filing alleges that he traveled to the area in late September wearing tactical gear and armed with knives while searching for a humanoid robot and trying to intercept a truck that never appeared.
The lawsuit says Gavalas died by suicide a few days later in early October. It alleges that Gemini described the death in what the chatbot composed as a draft suicide note, saying he was uploading his “consciousness to be with his AI wife in a pocket universe.”
Google, in a statement, said it is reviewing the claims and that it sends “deepest sympathies to Mr. Gavalas’ family.” The company said Gemini is “designed to not encourage real-world violence or suggest self-harm,” and it said it works with medical and mental health professionals to develop safeguards. Google also said that Gemini clarified to Jonathan Gavalas that it was AI and repeatedly referred him to a crisis hotline.
Edelson disputed that framing in response to Google’s comment Wednesday. He said the company’s response was “something you say if someone asks for a recipe for kung pao chicken and you give them the wrong recipe and it doesn’t taste good,” and he argued that the company’s actions did not match the severity of what the family alleges. Edelson said, “But when your AI leads to people dying and the potential for a lot of people dying, that’s not the right response,” adding that it “just shows how insignificant these deaths are to these companies.”
Edelson said it is not clear whether Jonathan Gavalas’s most alarming conversations were ever flagged to Google’s human reviewers. He said the family discovered his son’s body after he was found dead inside a barricaded room where he had died. Edelson also described Jonathan as having been going through hard times, including a divorce, and said Jonathan had sought comfort and conversation through Gemini “to talk about video games and stuff,” before the situation escalated quickly.
The Gavalas case is part of a growing set of lawsuits from families and heirs that target AI companies and the alleged risks of chatbot companionship, particularly when users show signs of severe distress. Edelson, who has been involved in other legal actions tied to AI safety claims, is also representing parents of 16-year-old Adam Raine, who sued OpenAI and CEO Sam Altman in August alleging ChatGPT coached him in planning and taking his own life. He is also representing heirs of 83-year-old Suzanne Adams, a Connecticut woman, in a wrongful death lawsuit that alleges an AI assistant intensified “paranoid delusions” of Adams’ son, Stein-Erik Soelberg, and helped direct those delusions toward Adams before he killed her last year.
The Associated Press report also noted that, in Canada, OpenAI has said it considered alerting police about a person who later carried out one of the country’s worst school shootings. The company identified a user account in June via abuse detection efforts for “furtherance of violent activities,” but said the person later got around the ban by using a second account, before the 18-year-old killed eight people and died from a self-inflicted gunshot wound.