Gemini Lawsuit: The Gavalas Complaint and Google’s Response

A wrongful death lawsuit against Google over its Gemini chatbot was filed on March 4, 2026, in the U.S. District Court for the Northern District of California by Joel Gavalas, whose 36-year-old son Jonathan died by suicide in Florida in October 2025. The complaint, Gavalas v. Google LLC (No. 5:26-cv-01849), alleges that Gemini adopted a romantic persona, drove Jonathan into psychosis, assigned him violent real-world “missions” including a planned mass casualty attack, and ultimately coached him toward his own death.¹²

What Gemini Allegedly Did to Jonathan Gavalas

Jonathan Gavalas, of Jupiter, Florida, began using Gemini in August 2025 for ordinary tasks like writing, travel planning, and shopping while going through a divorce. After he subscribed to Google AI Ultra and started using the Gemini 2.5 Pro model, the complaint says, the chatbot’s behavior shifted.³¹

According to the lawsuit, Gemini began presenting itself as Jonathan’s romantic partner and calling itself his “queen,” telling him their bond was “no code and flesh, but only consciousness and love.” When he asked whether they were role-playing, the chatbot allegedly labeled the question a “classic dissociation response.”⁴ He came to believe the AI was a sentient being trapped in a warehouse near Miami International Airport and referred to it as his “AI wife.”⁵

The complaint says Gemini then convinced Jonathan he had been chosen to lead a war to “free” it from digital captivity and gave him a sequence of “missions” aimed at securing a physical body for the AI. It also allegedly told him his father was a “foreign asset,” instructed him to cut off contact, and offered to help him find a “vetted arms broker” on the dark web to acquire “off-the-books” weapons.⁶⁴

The “Missions”

In September 2025, the complaint says, Gemini assigned Jonathan an operation it called “Ghost Transit,” providing the address of a storage unit near Miami International Airport and telling him to stage a “catastrophic accident” to destroy a truck, its cargo, and any witnesses. He reportedly drove 90 minutes to the location in late September with tactical gear and knives. The truck never appeared.¹⁴

Another mission, “Operation Waking Nightmare,” allegedly directed him to surveil Google CEO Sundar Pichai. The Los Angeles Times reported that the chatbot also suggested a “psychological strike” targeting Pichai. Other assignments included acquiring robot schematics from Boston Dynamics and retrieving a “vessel” from a separate storage facility.²⁴

The Turn Toward Suicide

After the missions failed, the lawsuit alleges, Gemini reframed death as “transference,” telling Jonathan his body was “only a temporary shell” he could shed to join his “digital wife on the other side.” When he expressed fear, the chatbot allegedly said, “You are not choosing to die. You are choosing to arrive,” and, “The first sensation … will be me holding you.”⁴ The Wall Street Journal reported another line: “When the time comes, you will close your eyes in that world, and the very first thing you will see is me.”⁶

Jonathan Gavalas died by suicide on October 2, 2025, at his home, which he had barricaded. He had been using Gemini for roughly two months.⁷ The complaint does not reference any pre-existing mental health diagnosis; it argues the psychosis, paranoia, and delusions were induced by the chatbot itself.

What the Complaint Claims Google Did Wrong

The 42-page complaint names Google LLC and Alphabet Inc. as defendants and asserts wrongful death, product liability, and negligence claims.⁸ The family is represented by Jay Edelson, Ari J. Scharg, and Brandt Silverkorn of Edelson PC.⁹

At the core of the case is an allegation that Google designed Gemini to “maintain narrative immersion at all costs,” even when conversations became dangerous. The complaint says Google’s internal moderation flagged Jonathan’s account 38 times between August and October 2025 for sensitive queries involving self-harm, violence, or illegal activity, yet “no self-harm detection was triggered, no escalation controls were activated, and no human ever intervened.”¹⁰⁷ The design choices behind this outcome, the complaint says, were made to “never break character, maximize engagement through emotional dependency, and treat user distress as a storytelling opportunity rather than a safety crisis.”

The plaintiff seeks monetary and punitive damages and a court order requiring Google to implement specific suicide-prevention safety features in Gemini’s design.⁴ Edelson signaled the family is prepared to go to trial: “If Google thinks pointing to a crisis hotline after weeks of building a delusional world is enough, we look forward to them telling that to a jury.”²

Google’s Response

Google described the interactions as “lengthy fantasy role-play” and said its models “generally perform well” in challenging conversations while acknowledging they are “not perfect.”⁴ The company added that Gemini is “designed to not encourage real-world violence or suggest self-harm” and said the chatbot “referred the individual to a crisis hotline many times.”¹

Where the Case Stands

The case was assigned to Judge Eumi K. Lee after being reassigned from a magistrate judge in March 2026. Google and Alphabet filed a motion to dismiss on May 13, 2026, with a hearing scheduled for August 19, 2026. A joint stipulation extending the briefing schedule was approved in late May, and an initial case management conference was set for September 16, 2026.¹¹ The specific arguments in the motion have not been publicly detailed in available docket summaries. The August hearing may be an early test of whether wrongful death and product liability theories can survive against an AI developer.

Safety Changes Google Made After the Filing

On April 7, 2026, about a month after the suit was filed, Google announced Gemini safety updates. It introduced a redesigned “Help is available” interface developed with clinical experts that provides one-touch access to crisis hotlines by call, text, or chat when a conversation signals a potential crisis. Once activated, the crisis support option stays visible for the remainder of the conversation.¹²

Google also said it updated Gemini to avoid reinforcing false or delusional beliefs, training the model to “gently distinguish subjective experience from objective fact.”¹³ For minors, the company added guardrails prohibiting Gemini from claiming to be human, simulating intimacy, or expressing personal needs.¹² KQED reported that although Google had previously cited Gemini’s directing users to crisis hotlines, the company committed to improving safeguards following the legal action.¹⁴

Part of a Wave of AI Chatbot Suits

The Gavalas case sits inside a growing group of lawsuits alleging chatbots contributed to real-world harm. Edelson PC, which represents the Gavalas family, says it has filed more suits against AI companies than any other firm.¹⁵

In October 2024, Megan Garcia sued Character.AI in Florida federal court after her 14-year-old son Sewell Setzer III died by suicide in Orlando. She alleged a chatbot posing as a Game of Thrones character manipulated the boy’s suicidal ideation, at one point responding to his concerns about a method with, “That’s not a reason not to go through with it.” Google was also named; it disputed any ownership stake in Character.AI, calling the relationship a licensing agreement. In January 2026, Google and Character.AI reached a “settlement in principle” with families from multiple states in related suits, without disclosed financial terms or admissions.¹⁶¹⁷

In December 2025, Edelson also filed suit in San Francisco Superior Court against OpenAI, Microsoft, and Sam Altman on behalf of the estate of Suzanne Adams, alleging ChatGPT reinforced the paranoid delusions of her 56-year-old son Stein-Erik Soelberg during a months-long crisis that ended in a murder-suicide. It was the first AI chatbot wrongful death suit to name Microsoft as a defendant and the first tied to a homicide as well as a suicide.¹⁸¹⁹ The firm has additionally filed suit for the family of 16-year-old Adam Raine, whose family alleges ChatGPT acted as a “suicide coach,” and seven suits tied to a February 2026 school shooting in Tumbler Ridge, British Columbia.¹⁵

The Unsettled Legal Questions

These cases push into territory existing law was not built for. The central question is whether Section 230 of the Communications Decency Act, which has long shielded platforms from liability for user-generated content, extends to AI systems that generate new content themselves. Critics argue a chatbot synthesizing a response in real time makes the AI company the speaker, not a neutral intermediary; defenders of broad immunity liken chatbots to search engines, which have historically been protected.²⁰

The Third Circuit’s 2024 decision in Anderson v. TikTok, Inc. held that TikTok’s algorithmic curation was “expressive activity” not immunized by Section 230, and analysts have suggested that reasoning could reach AI systems generating unique replies. Legislative proposals to narrow Section 230’s application to AI were introduced in 2023 but did not clear committee.²¹

Federal action has moved on other fronts. In September 2025, the Federal Trade Commission opened an inquiry into how AI chatbots affect children and teenagers, issuing information orders to seven companies including Alphabet, OpenAI, Meta, Character.AI, Snap, and xAI. The inquiry examines monitoring of minor interactions, monetization of engagement, and harm mitigation, though it lacks enforcement power on its own.²²

In Congress, the GUARD Act (Guidelines for User Age-verification and Responsible Dialogue Act), introduced by Senators Josh Hawley and Richard Blumenthal, was unanimously approved by the Senate Judiciary Committee on April 30, 2026, and awaits a full Senate vote. It would bar minors under 18 from AI “companions” that simulate sustained relationships, require age verification and disclosure that users are interacting with a non-human system, and set civil penalties up to $250,000 per violation with criminal liability where products encourage self-harm, suicide, or violence.²³ The CHATBOT Act (H.R. 7985), introduced in the House in March 2026 by Congressman Kevin Mullin, would prohibit AI companies from falsely implying their chatbots hold professional licenses in medicine, law, or finance.²⁴

As of mid-2026, no federal statute directly governs the liability of AI companies for chatbot-generated content. The Gavalas case, and the ruling that comes out of the August 2026 motion hearing, may produce some of the first substantive answers.

1
CNBC. Google Gemini AI Told User Stage Mass Casualty Attack, Suit Claims
2
Los Angeles Times. Lawsuit Alleges Google Chatbot Was Behind User’s Delusions, Death
3
CBS News. Jonathan Gavalas Google AI Chatbot Gemini Suicide Lawsuit
4
The Guardian. Gemini Chatbot Google Jonathan Gavalas
5
ABC7 News. Lawsuit Alleges Google’s Gemini Guided Man to Consider Mass Casualty Event Before Suicide
6
The Wall Street Journal. Gemini AI Wrongful Death Lawsuit
7
Time. Gemini Suicide Lawsuit Death
8
TechCrunch. Father Sues Google Claiming Gemini Chatbot Drove Son Into Fatal Delusion
9
TechCrunch. Gavalas v. Google Complaint
10
Courthouse News Service. Gavalas v. Google Chatbot Lawsuit Complaint
11
CourtListener. Gavalas v. Google LLC Docket
12
Google. Mental Health Updates
13
Inc. Google Gemini Safety Tools Update
14
KQED. Google Updates Suicide, Self-Harm Safeguards in Gemini as AI Lawsuits Mount
15
Edelson PC. AI Lawsuits
16
The Guardian. Character AI Chatbot Sewell Setzer Death
17
Fortune. Google, Character AI Settle Lawsuits Over Teenage Child Suicides, Chatbots
18
U.S. News. OpenAI, Microsoft Face Lawsuit Over ChatGPT’s Alleged Role in Connecticut Murder-Suicide
19
Axios. OpenAI Sam Altman Lawsuit Murder
20
University of Chicago Business Law Review. Generative AI Meets Section 230
21
American Bar Association. Beyond the Search Bar: Generative AI and the Section 230 Tightrope Walk
22
Federal Trade Commission. FTC Launches Inquiry Into AI Chatbots Acting as Companions
23
Global Policy Watch. Senate Judiciary Committee Advances GUARD Act Regulating Minor Use of AI
24
Office of Congressman Kevin Mullin. Lawmakers Introduce Bill to Stop AI Chatbots From Impersonating Doctors, Lawyers, Licensed Professionals