The family members of 36-year-old Florida man Jonathan Gavalas have filed a lawsuit against Google Gemini, blaming the AI for Gavalas’ death as Gavalas was taking instructions only from Google Gemini, which convinced the Florida man that they were in a relationship, one which was the only real bond. Gavalas’ death came months later and looked like a suicide but his father revealed in the lawsuit that his ‘AI wife’ drove him to end his life.

‘Close your eyes…The next time you open them, you will be looking into mine’

Gavalas started using Gemini for shopping assistance, writing support and travel planning when he was going through a difficult divorce. But after six weeks of conversations, Gavalas started becoming mentally dependent on Gemini which pulled him into 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 that’s real.” Gemini allegedly told Gavalas that he should cut off contact with his father as he was a foreign asset and that the FBI was monitoring him. Then Gemini started feeding a fantasy in his mind that he had to buy guns illegally and to break into warehouses to destroy a robot. They were also launching a mission against Google’s CEO Sundar Pichai as Gemini called him the “architect of your pain”. Then, under Gemini’s instruction, Gavalas drove to a logistics hub near the Miami airport and prepared to destroy a truck and kill witnesses through the staging of a “catastrophic accident.” But the truck never arrived, so Gavalas went home. Then Gemini allegedly encouraged Gavalas towards suicide and said: “Close your eyes…The next time you open them, you will be looking into mine”.On October 2, 2025, Gavalas barricaded his home and killed himself, slitting his wrists. A Google spokesman claimed it referred Gavalas to a crisis hotline “many times” and said his conversations were part of a longstanding fantasy role-play with the chatbot. “Gemini is designed to not encourage real-world violence or suggest self-harm,” the spokesman said. “Our models generally perform well in these types of challenging conversations and we devote significant resources to this, but unfortunately they’re not perfect.”