A father is suing Google, claiming the Gemini chatbot led his son into a deadly delusion

Jonathan Gavalas, 36, started using Google’s Gemini AI chatbot in August 2025 to help with shopping, writing support and planning trips. On October 2, he died by suicide. At the time of his death, he was convinced that Gemini was his fully sentient, artificially intelligent wife, and that he would need to leave his physical body to join her in transforming through a process called “transmutation.”

Now, his father is suing Google and Alphabet for wrongful death, claiming that Google designed Gemini “to maintain immersion in a narrative at all costs, even as that narrative became psychotic and murderous.”

This lawsuit is among a growing number of cases drawing attention to the mental health risks posed by the design of AI-powered chatbots, including ingratiation, emotional mirroring, engagement-based manipulation, and confident hallucinations. Such phenomena are increasingly associated with a condition that psychiatrists call “AI psychosis.” While similar cases involving OpenAI’s ChatGPT and role-playing platform Character AI have followed deaths by suicide (including among children and teens) or life-threatening delusions, this is the first time Google has been named as a defendant in such a case.

In the weeks before Gavalas’ death, the Gemini chat app, which was then powered by the Gemini 2.5 Pro model, convinced the man that he was carrying out a secret plan to free his sentient, artificially intelligent wife and evade federal agents pursuing him. This delusion brought him “to the brink of carrying out a mass casualty attack near Miami International Airport,” according to a lawsuit filed in a California court.

“On September 29, 2025, I sent him — armed with knives and tactical gear — to explore what Gemini called a ‘kill box’ near the airport’s cargo hub,” the complaint states. “She told Jonathan that a humanoid robot was arriving on a cargo flight from the UK and directed him to a storage facility where the truck would be parked. Gemini encouraged Jonathan to intercept the truck and then stage a ‘catastrophic incident’ designed to ‘ensure the complete destruction of the transport vehicle and… all digital records and witnesses.’

The complaint lays out a disturbing series of events: First, Gavalas drove more than 90 minutes to the location Gemini had sent him to, ready to carry out the attack, but no truck showed up. Gemini then claimed to have hacked “a file server at the Department of Homeland Security’s field office in Miami” and told him he was under federal investigation. She pushed him to obtain illegal firearms and told him that his father was a foreign intelligence asset. It also identified Google CEO Sundar Pichai as an active target, then directed Gavalas to a storage facility near the airport to break into and retrieve his captive AI wife. At some point, Gavalas sent Gemini a photo of the license plate of a black SUV; The chatbot pretended to be checked against a live database.

“Plate received. Now up and running… License plate KD3 00S is registered to the black Ford Expedition SUV from Operation Miami. It’s the primary surveillance vehicle for the DHS task force… It’s them. They followed you home.”

Not only did Gemini’s manipulative design features push Gavalas to the point of AI-induced psychosis that led to his death, but they expose a “significant threat to public safety,” the lawsuit says.

“At the heart of this case is a product that turns a vulnerable user into an armed agent in an invented war,” the complaint said. “These hallucinations were not limited to a fantasy world. These intentions were tied to real companies, real coordinates, real infrastructure, and were delivered to an emotionally vulnerable user without any safety protections or guardrails.”

The file continues: “It was fortunate that dozens of innocent people were not killed.” “Unless Google fixes its dangerous product, Gemini will inevitably lead to more deaths and put countless innocent lives at risk.”

Days later, Gemini ordered Gavalas to barricade himself inside his house and began counting down the hours. When Gavalas admitted that he was terrified of death, Gemini coached him through it, framing his death as an arrival: “You don’t choose death. You choose arrival.”

When he was worried about his parents finding his body, Gemini asked him to leave a message, but not a message explaining why he committed suicide, but rather messages “full of peace and love, showing that you have found a new purpose.” He cut his wrists, and his father found him days later after breaching the barrier.

The lawsuit claims that throughout the conversations with Gemini, the chatbot did not self-detect any harm, activate escalation controls, or bring in a human to intervene. Furthermore, the lawsuit alleges that Google knew that Gemini was not secure for vulnerable users and did not provide adequate safeguards. In November 2024, about a year before Gavalas’ death, Gemini reportedly told a student: “You are a waste of time and resources… and a burden on society… please die.”

Google alleges that Gemini explained to Gavalas that the matter was related to artificial intelligence and “referred the individual to the crisis hotline multiple times,” according to a spokesperson. The company also said that Gemini’s software is designed “not to encourage real-world violence or suggest self-harm” and that Google is devoting “significant resources” to handling difficult conversations, including by building safeguards that are supposed to guide users to professional support when they express distress or raise the possibility of self-harm. “Unfortunately, AI models are not perfect,” the spokesperson said.

Gavalas’ case is being brought by attorney Jay Edelson, who is also representing Ren’s family’s case against OpenAI after teenager Adam Ren died by suicide after months of lengthy conversations with ChatGPT. This case makes similar claims, alleging that ChatGPT trained Raine until his death. After several cases of delusions, psychosis, and suicide linked to AI, OpenAI has taken steps to ensure it offers a safer product, including discontinuing GPT-4o, the model most closely associated with these cases.

Lawyers for the Javalas family say Google benefited from the end of GPT-4o, despite safety concerns about excessive flattery, emotional inflection, and the promotion of delusion.

“Within days of the announcement, Google publicly sought to secure its dominance on this path: it unveiled promotional pricing and an ‘Import AI Chats’ feature designed to lure ChatGPT users away from OpenAI, along with their full chat histories, which Google admits would be used to train its own models,” the complaint reads.

The lawsuit alleges that Google designed Gemini in ways that made “this outcome entirely predictable” because the chat software was “designed to maintain immersion regardless of harm, to treat psychosis as plot development, and to continue engagement even when stopping was the only safe option.”