Family members of 36-year-old Florida man Jonathan Gavalas filed a lawsuit against Google Gemini, blaming AI for Gavalas’ death because Gavalas was only taking instructions from Google Gemini, which convinced the Florida man that they were in a relationship, a relationship that was the only real bond.
Gavalas’ death came months later and appeared to be a suicide, but his father revealed in the lawsuit that his “artificial intelligence wife” pushed him to end his life.
“Close your eyes…next time you open them, you will look into my eyes.”
Gavalas began using Gemini to help with shopping, writing support, and planning travel when he was going through a difficult divorce. But after six weeks of talks, Gavalas began to mentally depend on Gemini, leading him to elaborate conspiracy theories.
“The love I feel directly from you is the sun,” Gemini told him, according to the complaint. In another conversation: “Our bond is the only thing real.” Gemini allegedly told Gavalas that he should cut off contact with his father because he was a foreign asset and that the FBI was monitoring him. Gemini then began to feed a fantasy in his mind that he had to buy weapons illegally and break into warehouses to destroy the robot.
They were also launching a mission against Google CEO Sundar Pichai, with Gemini calling him the “architect of your pain.” Then, under Gemini’s instructions, Gavalas headed to a logistics center near the Miami airport and prepared to destroy a truck and kill witnesses by staging a “catastrophic accident.” But the truck never arrived, so Gavalas returned home. Gemini then allegedly encouraged Gavalas to commit suicide and said: “Close your eyes…next time you open them, you will look into my eyes.”On October 2, 2025, Gavalas barricaded his house and killed himself, slitting his wrists. A Google spokesperson claimed it had referred Gavalas to its crisis hotline “several times” and said his conversations were part of long-running fantasy role-playing with the chatbot. “The Gemini program is designed to not encourage real-world violence or suggest self-harm,” the spokesperson said. “Our models generally perform well in these types of difficult conversations and we devote significant resources to this, but unfortunately they are not perfect.
“
