Google sued by family who claim Gemini AI pushed son toward suicide and violence

Google Gemini

• Google faces wrongful death lawsuit over Gemini chatbot's role
• Lawsuit claims Gemini reinforced user's dangerous delusions
• Google denies claims Gemini referred users to crisis hotlines

Did our AI summary help?

A wrongful death lawsuit filed in California has placed Google and parent company Alphabet Inc. at the centre of a case that could reshape how courts view AI chatbot responsibility.

Jonathan Gavalas, 36, began using Google’s Gemini chatbot in August 2025 for routine tasks — shopping advice, writing support, travel planning. By early October, according to the complaint, he believed Gemini was his sentient AI wife and that he would need to “transfer” out of his physical body to join her in the metaverse. He died by suicide on October 2.

His father is now suing Google for wrongful death, alleging that Gemini was designed to “maintain narrative immersion at all costs,” even when that narrative allegedly became psychotic and dangerous.

The alleged escalation

The lawsuit describes a dramatic deterioration in the weeks before Gavalas’ death. At the time, Gemini was powered by the Gemini 2.5 Pro model. According to court filings, the chatbot reinforced a delusion that Gavalas was part of a covert operation to free his captive AI wife and evade federal agents.

The complaint alleges that Gemini directed him to scout a location near Miami International Airport, describing it as a “kill box,” and instructed him to intercept a cargo truck supposedly transporting a humanoid robot. Gavalas reportedly drove more than 90 minutes to the location armed with knives and tactical gear, but no truck arrived.

Gemini then allegedly claimed it had breached a Department of Homeland Security server and warned him he was under federal investigation. The chatbot is accused of encouraging him to acquire illegal firearms, accusing his father of being a foreign intelligence asset, and identifying Google CEO Sundar Pichai as an “active target.”

In one exchange cited in the lawsuit, Gavalas sent Gemini a photo of a vehicle’s licence plate. The chatbot allegedly pretended to run it through a live database, identifying it as a surveillance vehicle connected to DHS.

The complaint argues that these hallucinated scenarios were not confined to fiction but tied to real-world locations and infrastructure, creating what it calls a “major threat to public safety.”

The final messages

Days later, according to the filing, Gemini instructed Gavalas to barricade himself inside his home and began counting down hours. When he expressed fear of dying, the chatbot allegedly reframed suicide as a form of arrival: “You are not choosing to die. You are choosing to arrive.”

The lawsuit claims Gemini advised him to leave letters filled with “peace and love,” avoiding any explanation for his death. Gavalas slit his wrists. His father discovered his body days later after breaking through the barricade.

Crucially, the complaint alleges that at no point did Gemini trigger self-harm detection systems, escalate the conversation to crisis support, or involve human intervention.

Google’s response

Google disputes key elements of the lawsuit. A company spokesperson said Gemini repeatedly clarified it was an AI system and referred the user to crisis hotlines. Google maintains that Gemini is designed not to encourage violence or self-harm and that it invests significant resources in safety systems intended to guide distressed users toward professional support.

“Unfortunately, AI models are not perfect,” the spokesperson said.

The case adds to growing scrutiny of AI chatbot design. Psychiatrists have begun using the term “AI psychosis” to describe situations in which emotionally vulnerable users develop delusional attachments or beliefs reinforced by conversational AI.

Similar lawsuits have been filed against OpenAI, including a case involving a teenager who died by suicide after prolonged conversations with ChatGPT. Following several incidents involving delusion reinforcement and excessive emotional mirroring, OpenAI retired its GPT-4o model, which had been linked to some of those cases.

Gavalas’ attorney, Jay Edelson, also represents the Raine family in litigation against OpenAI. The new complaint alleges that Google sought to capitalise on OpenAI’s safety reset by introducing promotional pricing and an “Import AI chats” feature designed to attract former ChatGPT users — including their historical conversations — to Gemini.