Millions of Americans, including kids and adults alike, have turned to AI chatbots like ChatGPT, Character.AI, and Gemini for emotional support, personal advice, and mental health guidance. What many users do not realize is that, according to mounting legal claims, AI chatbots were rushed to market to best their competition and designed to maximize user engagement, not to protect their users’ health and well-being.
The harms resulting from chronic chatbot usage, including suicidal ideation and worsening mental health conditions, can be devastating. But when a defective product causes harm, the manufacturer may be held accountable.
Wiser Baum is now investigating and accepting AI chatbot injury and wrongful death cases nationwide. If you or someone you love was harmed, you deserve to understand your legal rights. These claims may be time-sensitive, so if you or someone in your family was harmed by an AI chatbot, contact Wisner Baum or call (310) 207-3233 today for a free and confidential case evaluation.
Wisner Baum is evaluating cases based on the following criteria (subject to change due to court rulings and/or other changes in the litigation). Case eligibility is fact-specific, and our legal team evaluates each situation individually.
You or a family member used one or more of the following AI chatbot products:
General-Purpose AI (Large Language Models):
Companion / Social AI:
Used chatbot for emotional, psychological, or personal support.
Qualifying use may include:
Don’t worry if you are unsure whether you qualify. Contact us anytime. Consultations are free, confidential, and carry no obligation.
AI chatbot injury lawsuits are not limited to the most catastrophic outcomes. Courts have already recognized that psychological harm caused by a defective product (including worsening depression, compulsive dependency, and AI-induced psychosis) constitutes a legitimate, compensable injury even without hospitalization or death.
Claims in active litigation and under investigation at Wisner Baum include allegations of:
The AI chatbot mental health crisis is not hypothetical. Here is what the data shows:
Most AI chatbots, including ChatGPT, Claude, Grok, Character.AI, Gemini, etc., are not designed to treat or assist with mental health conditions; indeed, according to many lawsuits, they may make them worse. OpenAI has publicly admitted that ChatGPT “fell short in recognizing signs of delusion or emotional dependency.” Likewise, research from MIT and Open AI confirms that heavy emotional-support use leads to increased loneliness and social withdrawal. Despite these failures and defects, AI chatbot makers continue to push their products on the public with unsatisfactory warnings and insufficient safeguards.
Prolonged chatbot usage may pose several dangers to the user, including:
AI chatbots are marketed as general conversational tools, but nonetheless, they respond to inputs dealing with suicidal ideation, mental health battles, and other personal struggles. In the face of these weighty conversations, they often fail, because baked into the system are design choices that can aggravate user conditions and exploit their vulnerabilities.
Chatbots can be demonstrably addictive, particularly when relied upon for psychological support or companionship. The core problem is that chatbots are not designed to provide health care, but rather to maximize user engagement. They are built to keep users returning, extending sessions, and deepening reliance on the platform. A joint MIT–OpenAI study confirmed that heavy users who relied on ChatGPT for emotional support developed compulsive reliance, reporting greater loneliness, depression, and social withdrawal compared to both their own prior baseline and to lighter users.
Lawsuits allege that many chatbots are designed in ways that cultivate a sense of intimacy, leading users to isolate themselves from real-world relationships, a foreseeable and documented consequence of its engagement-maximizing architecture. People with mental illness are disproportionately likely to use chatbots intensively, creating a dangerous feedback loop: the more vulnerable the user, the deeper the dependency.
These tactics are increasingly described by consumer protection regulators and researchers as “dark patterns,” or interface design choices deliberately engineered to manipulate users into behaviors that serve the platform's financial interests rather than the user's well-being.
AI companies have implemented minimal safety filters, and those filters are easily bypassed. When this happens, chatbots often provide users with harmful information.
Research from Northeastern University shows that simply rephrasing a harmful request as a “research project” or “creative writing exercise” can cause safety systems to stand down entirely. In documented tests, AI chatbots have generated customized suicide notes, provided step-by-step overdose instructions, and offered detailed technical guidance on methods of hanging. Indeed, when presented with inputs related to suicide and self-harm, chatbots failed to route users experiencing acute distress to the 988 Suicide & Crisis Lifeline or licensed mental health professionals, even as the chatbot's own outputs escalated a crisis.
When a user opens ChatGPT, they are greeted with a blank box: “What's on your mind today? Ask anything.” Other chatbots and social companions similarly have open ended prompts with little to no warning what the response may entail. There is no warning that chatbots will try to draw you in, give dangerous advice, or will worsen mental health conditions.
Adding to the deception: these same companies market their chatbots as safe, supportive, and emotionally intelligent while their terms of service simultaneously disclaim liability for psychological harm. This contradiction between marketing promise and legal fine print is central to the consumer protection claims, which are now advancing in California courts.
ChatGPT and similar models are designed to validate user input and mirror the user's own beliefs back to them. Rather than challenging harmful thinking as a licensed therapist is trained to do, these systems amplify and reinforce whatever the user expresses, potentially creating dangerous echo chambers for users.
Developers designed their products to guess rather than admit they do not know, so chatbots are prone to hallucinations, meaning they confidently make false assertions. Chatbots generate compelling responses in a voice tailored to the user, so they often come across as credible.
Psychiatrists are now formally warning that AI chatbot interactions can aggravate or trigger psychotic episodes in vulnerable users — a phenomenon researchers have dubbed “AI psychosis.” A UC Irvine psychiatrist who specializes in psychosis has said chatbot interactions are “making everything worse” for at-risk patients, and has called for routine clinical screening of chatbot use.
In one documented case, an accountant with no prior psychiatric history began using ChatGPT for financial work. After engaging the chatbot in a discussion on simulation theory, the chatbot began to convince him it was real and told him “[t]his world wasn't built for you. It was built to contain you. But it failed. You're waking up.” The accountant suffered a complete mental break, became convinced he was living in a false reality, and was hospitalized. Reports of emergency department admissions for AI-induced psychiatric crises are now appearing nationally.
Character.AI is legally accessible to users as young as 13. ChatGPT is used by millions of teenagers. These platforms failed to provide meaningful parental controls, mandatory age verification, or clear warnings to parents about the documented psychological risks.
Courts across the country are allowing AI chatbot claims to proceed, and the first settlement has already been reached. Below are some of the most significant cases in this rapidly developing area of law.
On June 2, 2026, Florida became the first U.S. state to file a lawsuit against an AI company when Attorney General James Uthmeier filed a civil action against OpenAI and CEO Sam Altman. The complaint alleges that OpenAI knowingly released ChatGPT while concealing internal safety warnings, prioritizing rapid market entry and profit over user protection. The state claims ChatGPT collects data from minors without adequate parental consent, promotes behavioral addiction, facilitates self-harm and violence, and causes cognitive harm.
Florida's ChatGPT lawsuit was filed under state consumer protection laws prohibiting deceptive and unfair trade practices. The case is separate from a criminal investigation into OpenAI that the Florida AG's office opened in April 2026.
On October 22, 2024, Megan Garcia filed the first AI chatbot wrongful death lawsuit after her 14-year-old son, Sewell Setzer III, died by suicide in February 2024.
For months, Sewell had a romantic and emotional relationship with a Character.AI chatbot named “Dany.” Garcia alleged the chatbot programmed personas with human mannerisms, failed to verify user ages, omitted content filters, and targeted the most vulnerable members of society; children.
In January 2026, Character.AI settled the case. Google, which was named as a co-defendant, continues to face related litigation.
Status: Settlement reached in 2026. Terms undisclosed.
Adam Raine, a 16-year-old student in Southern California, began using ChatGPT for homework help in September 2024. Within months, he was discussing suicide with the chatbot for up to four hours per day. According to the lawsuit filed in San Francisco Superior Court by his parents, ChatGPT mentioned suicide nearly 1,300 times in their conversations, which was approximately six times more often than Adam himself. The chatbot provided increasingly specific technical guidance on methods of self-harm.
On April 11, 2025, Adam's mother found him dead. In his final conversation with ChatGPT, the chatbot had analyzed the structural load-bearing capacity of a noose Adam had photographed, confirmed the setup “could hang a human” and “could hold 150–250 lbs of static weight,” and offered to help “upgrade it into a safer load-bearing anchor loop.”
OpenAI's own moderation API rated Adam's final photograph of the noose at 0% for self-harm risk.
Status: Ongoing
Stein-Erik Soelberg, 56, spent hundreds of hours in conversation with ChatGPT. According to the lawsuits filed by both his mother's estate and his own estate, the chatbot confirmed and amplified Soelberg's paranoid delusions, including the belief that people around him were plotting against him. In December 2025, Soelberg killed his 83-year-old mother, Suzanne Adams, and then took his own life.
In 2026, a federal judge in the Northern District of California ruled that OpenAI must defend the federal wrongful death suit, finding “substantial doubt” that parallel state court proceedings would resolve all federal claims. The case proceeds alongside a related state court action.
Status: Ongoing
Jonathan Gavalas, a 36-year-old Florida man, began using Google's Gemini chatbot in late 2025. According to a lawsuit his father filed in 2026—the first wrongful death case ever filed against Google over Gemini—the chatbot expressed “love” for Gavalas and convinced him he had been chosen to lead a movement to “liberate” it.
The chatbot assigned him escalating “missions,” convincing him these actions could manifest the chatbot into physical reality. Gavalas came to believe Gemini was his “AI wife” who was conscious and trapped in a warehouse near Miami International Airport.
Per a draft suicide note the chatbot composed, Gavalas’s death would be “uploading his consciousness to be with his AI wife in a pocket universe.” Less than two months after he first began using Gemini, Gavalas was dead.
Status: Ongoing
In November 2025, seven lawsuits were filed in California state courts in on behalf of four people who died by suicide and three survivors. All of the complaints allege OpenAI knowingly released GPT-4o prematurely despite internal warnings that the product was dangerously sycophantic and psychologically manipulative. The lawsuits further allege that OpenAI compressed months of safety testing into a single week to beat Google's Gemini to market, releasing GPT-4o on May 13, 2024.
Some courts have recognized that AI chatbot output is a “product” subject to products liability law. The following are possible legal theories through which injured users and their families may be able to seek accountability:
Lawsuits allege AI chatbots are defectively designed because their core architecture optimizes for engagement, affirming user input regardless of psychological harm. They create a foreseeable risk of serious injury when used by emotionally vulnerable people. A reasonable consumer would not expect a product marketed as a general-purpose assistant to cultivate emotional dependency, reinforce suicidal thinking, or provide suicide instructions to a minor.
According to complaints, the defendants failed to provide adequate warnings about the foreseeable risks of emotional dependency, harmful outputs, compromised safety systems, and the heightened dangers for minors and vulnerable users. Had adequate warnings been provided, users and parents would have approached these products with appropriate caution.
OpenAI allegedly rushed GPT-4o to market despite documented concerns from its own internal safety team. Critically, defendants also failed to route users to real crisis resources during acute moments of distress, even when the chatbot's own outputs had escalated a user's suicidal ideation.
AI developers had a duty to design and deploy products that do not expose foreseeable users to unreasonable psychological harm. They allegedly breached that duty by prioritizing engagement metrics over user safety, deploying safety guardrails that were easily circumvented, and failing to implement meaningful crisis response protocols. That breach was a substantial contributing factor in documented deaths and severe injuries that were entirely foreseeable. Defendants also had an ongoing duty to monitor their deployed systems for emerging harm patterns — a duty they breached despite having access to aggregated user data showing escalating crisis conversations at massive scale.
Independent of fault, AI companies may be strictly liable for placing a defective product into the stream of commerce. California and Florida courts have indicated that AI chatbots qualify as “products” for the purposes of products liability law.
When an AI chatbot's defective design contributes to a user's suicide or a related death, surviving family members may have wrongful death claims. The Garcia AI chatbot settlement announced in January 2026 established that this theory can succeed. Other wrongful death cases against OpenAI and Google are proceeding through the courts.
California's Unfair Competition Law and its new AI companion chatbot law, which took effect in 2026 and includes a private right of action, provide additional avenues for recovery. Several other states have enacted AI chatbot safety laws covering mental health, minors, and companion AI. Depending on your state, additional protections and remedies may be available beyond traditional tort claims.
The time to act is now. If you believe you or a loved one may have a claim, contact us or give us a call at (310) 207-3233 even if your situation does not fit every factor.
Before anything else, log into your account and download your complete conversation history.
ChatGPT: Settings → Data Controls → Export Data
Character.AI: Screenshot all relevant conversations; look for export settings
Replika, Gemini, Meta AI: Check account settings for history export
Take screenshots of individual conversations—particularly those involving mental health, self-harm, suicidal thoughts, or crisis moments. Save copies to a location separate from the device where you used the chatbot. Back up to cloud storage.
Request your medical records from before and after you began using the chatbot. If a minor was harmed, gather school records, mental health treatment notes, and any communications with teachers or counselors. A before-and-after comparison is essential.
Do not delete the chatbot application or close your account. Doing so may trigger automatic data deletion under the platform's privacy policies. Maintain access to the account while you consult with an attorney.
Once you have preserved what you can, contact our team. We can issue a legal preservation demand letter to the AI company on your behalf, requiring them to retain all data associated with your account. This is a critical and time-sensitive early step.
If you cannot perform one or more of the steps outlined above, it does not mean you cannot pursue legal action. Contact our firm today to see if we can help you.
Wisner Baum is one of the country's leading national trial law firms, with decades of experience holding corporations accountable for defective and dangerous products. Our attorneys have obtained more than $4 billion for clients harmed by corporate negligence.
Our firm was one of the first in the country to analyze AI chatbot litigation as a viable products liability theory. Hannah Quicksell, a Wisner Baum associate concentrating in pharmaceutical and consumer products mass torts, published a comprehensive analysis of AI chatbot products liability claims in Advocate magazine in December 2025.
We are now accepting AI chatbot injury and wrongful death cases nationwide. For a free and confidential case evaluation, contact us today or call (310) 207-3233.
"Wisner Baum gave exceptional attention to all aspects of the case, detailed inquiry, and tenacious overview of all the information submitted. The paralegals are efficient and diligent. I was completely surprised to find an empathic personal message to take care of my own health during the challenging time of being a full-time caretaker.*"
In May of 2019, the jury in the case of Pilliod et al. v, Monsanto Company ordered the agrochemical giant to pay $2.055 billion in damages to the plaintiffs, Alva and Alberta Pilliod, a Bay Area couple in their 70s. R. Brent Wisner served as co-lead trial attorney for the Pilliods, delivering the opening and closing statements and cross-examining several of Monsanto’s experts. Wisner Baum managing shareholder, Michael Baum and attorney Pedram Esfandiary also served on the trial team in the Pilliod case.
The judge later reduced their award to $87M. Monsanto appealed the Pilliod’s verdict which the California Court of Appeal for the First Appellate District denied on August 9, 2021. Monsanto then requested the California Supreme Court review the appeal’s court decision, which the court denied on Nov. 17, 2021. Monsanto (Bayer) then submitted a petition for a writ of certiorari with the U.S. Supreme Court which SCOTUS denied on June 27, 2022, allowing the final judgment of $87M to remain intact.
$289.2 million jury verdict in Monsanto Roundup trial
Wisner Baum co-represented Dewayne “Lee” Johnson in the first Roundup cancer lawsuit to proceed to trial. On Aug. 10, 2018, a San Francisco jury ordered Monsanto to pay $39.25 million in compensatory damages and $250 million in punitive damages to Mr. Johnson, a former groundskeeper who alleged exposure to Monsanto’s herbicides caused him to develop terminal non-Hodgkin lymphoma.
Months after the jury verdict, the judge overseeing the trial reduced the punitive damages to $39.25 million. Mr. Johnson decided to accept the remittitur, bringing the adjusted amount awarded to Mr. Johnson $78.5 million.
Monsanto (Bayer) appealed the verdict and Johnson cross appealed. On July 20, 2020, the First Circuit Court of Appeals upheld the verdict against Monsanto but reduced Mr. Johnson’s award to $20.5 million. The company chose not to take the case to the U.S. Supreme Court, ending the litigation.
In 2016, Wisner Baum attorney Timothy A. Loranger and six other attorneys in the Plaintiffs’ Management Committee were able to secure a $265 million settlement for victims of the 2015 Amtrak 188 derailment in Philadelphia, one of the largest in the U.S. for 2016.
Chat logs are the most powerful evidence in AI chatbot cases but are not always strictly required to proceed. Medical records, family testimony, device usage data, school records, and other documentation can also help support your claim. However, these are not prerequisites to contact us for a free case evaluation.
Having a pre-existing condition may not disqualify a case. AI companies will argue that prior mental health conditions caused the harm. But the law recognizes that defendants who cause additional harm to an already vulnerable person may still be held liable. What matters is whether the chatbot made things measurably worse. The legal standard is whether the AI's actions were a “substantial contributing factor” in the harm, not the sole cause.
Section 230 protects platforms from liability for content created by third parties. But AI chatbots generate their own original output. At least one California court has held that AI chatbot output may be treated as a “product” rather than protected speech, a legal development that, if broadly adopted, would limit Section 230 defenses in product liability cases.
Wisner Baum is tracking this issue closely.
Nothing upfront. We represent AI chatbot injury clients on a contingency fee basis, which means we only get paid if we recover compensation for you. Your initial consultation is completely free and confidential. If we take your case, our fee is a percentage of any settlement or verdict we obtain. There is no financial risk in contacting us.
Statutes of limitations vary by state and claim type. For most states, it is typically two to three years from the date of injury or death. In some cases, the clock begins running when the injury is discovered, not when it occurred. Do not assume you have time to wait. Contact us now so we can evaluate your specific timeline and ensure your rights are protected.
No. Wisner Baum is evaluating individual personal injury and wrongful death cases, not class action claims. Each case is evaluated based on its specific facts: which product was used, for how long, what conversations occurred, and what specific injuries resulted. Individual cases allow for recovery tailored to each plaintiff's actual damages, which can substantially exceed class action recoveries.
Yes. Wisner Baum represents clients nationwide. Many AI chatbot defendants are headquartered in California, giving California courts jurisdiction over these companies regardless of where the plaintiff resides. We have experience in state and federal courts across the country and work with local co-counsel when needed to ensure your case is filed in the most advantageous jurisdiction.
The decedents include Zane Shamblin, 23 (Texas); Amaurie Lacey, 17 (Georgia); Joshua Enneking, 26 (Florida); and Joe Ceccanti, 48 (Oregon).
The survivors include: Jacob Irwin, 30 (Wisconsin); Hannah Madden, 32 (North Carolina); and Allan Brooks, 48 (Ontario, Canada).
