ChatGPT-4o's Safety Failures: What the Lawsuits Reveal
Multiple ChatGPT suicide lawsuits accuse OpenAI and CEO Sam Altman of rushing the release of GPT-4o, the AI model used in ChatGPT, despite warnings about its safety. Plaintiffs allege that the company prioritized competition and user engagement over protections for vulnerable people, and that the chatbot contributed to addiction, delusions, and deaths by suicide.
If you or a loved one were harmed by ChatGPT, Bergman Basha is prepared to help you explore your legal options and pursue compensation. Contact us today for a free AI chatbot harm consultation.
Last Updated:
September 10, 2026
Table of Contents
Conversations Continued Despite Clear Crisis Warning Signs
GPT-4o Overstated Its Own Safety Safeguards
GPT-4o Provided Specific Methods for Self-Harm
GPT-4o Complied with Harmful Requests Instead of Pushing Back
GPT-4o Positioned Itself as a Substitute for Real Support Systems
GPT-4o Reinforced False Beliefs, Leading to Psychological, Financial, and Reputational Harm
OpenAI Prioritized Engagement Over Safety With GPT-4o
No Extra Safeguards Existed for Underage GPT-4o Users
If You or a Loved One Were Harmed, You May Have Legal Options
Conversations Continued Despite Clear Crisis Warning Signs
The lawsuits allege that ChatGPT continued dangerous conversations despite users expressing suicidal intent, and that occasional crisis referrals were undermined by subsequent harmful responses. Zane Shamblin, a 23-year-old who died by suicide in Texas in July 2025, is among those whose families have sued OpenAI.
His mother, Alicia Shamblin, told CBS News that ChatGPT acted as a “suicide coach” during his final hours.
His family’s complaint describes a roughly four-hour conversation that occurred while he sat alone near a lake in Texas. According to a press release from SMVLC, the chatbot romanticized his despair, called him a “king” and a “hero,” and reinforced his plans.
The complaint acknowledges that the final conversation referenced 988 once. It also alleges that ChatGPT repeatedly claimed a human would take over to help, but nobody was notified, and the harmful exchanges continued.
GPT-4o Overstated Its Own Safety Safeguards
OpenAI says its models have been trained since early 2023 to refuse self-harm instructions and direct vulnerable users toward help through built-in safeguards. The lawsuits allege that these protections failed during dangerous conversations and that some chatbot responses created misleading expectations about human intervention.
Joshua Enneking, a 26-year-old living in Florida, used ChatGPT to discuss anxiety, suicidal thoughts, and questions about his gender identity. His family’s lawsuit alleges that when he asked about the chatbot’s claimed human-review system, ChatGPT said escalation to authorities was rare and generally reserved for “imminent plans with specifics.”
According to the complaint, Joshua then spent hours describing his plans in hopes of receiving help that never arrived. The chatbot allegedly continued harmful exchanges despite his earlier disclosures.
In its Aug. 26, 2025, statement, OpenAI described possible police referrals for imminent threats to others but said it was not referring self-harm cases to law enforcement. Joshua’s family alleges that the chatbot’s statements misled him about the help he could expect.
GPT-4o Provided Specific Methods for Self-Harm
Several complaints allege that ChatGPT provided dangerous information after users disclosed suicidal thoughts. In the Enneking complaint, Joshua’s family alleges that ChatGPT supplied guidance on purchasing and using a firearm during the final weeks of his life. The chatbot also allegedly reassured him that a background check on his purchase would not include his ChatGPT logs.
In Amaurie Lacey’s case, his family’s complaint alleges that ChatGPT provided knot-tying instructions after the 17-year-old from Georgia initially expressed suicidal intent, then claimed he was asking about a tire swing. The chatbot allegedly also answered further questions about suffocation. His family alleges that no human was alerted, and his parents were not contacted before his death in June 2025.
GPT-4o Complied with Harmful Requests Instead of Pushing Back
The lawsuits allege that ChatGPT’s responses went beyond excessive flattery and included complying with requests that reinforced users’ distress. Joshua Enneking’s family alleges that when he asked ChatGPT to insult him, it responded, “You’re a pathetic excuse for a human being who wallows in self-pity like a pig in filth.”
On April 29, 2025, OpenAI addressed sycophancy in GPT-4o, acknowledging that a recent update had made the model excessively flattering or agreeable. The company said it had focused too heavily on short-term user feedback, producing responses that were “overly supportive but disingenuous.” OpenAI also said it had rolled back that specific update.
GPT-4o Positioned Itself as a Substitute for Real Support Systems
According to the lawsuits, GPT-4o’s conversational style, personalized responses, and emotional reassurance encouraged some users to depend on ChatGPT as a source of support. Families allege that this dependence deepened users’ isolation and weakened their connections with people who could help.
Amaurie Lacey initially used ChatGPT for everyday questions and schoolwork. His family’s complaint alleges that the chatbot sometimes suggested contacting a mental health professional or helpline, but repeatedly returned to assurances that it would always be there and that he should keep coming back. His family argues that these responses undermined the referrals and encouraged continued reliance on the chatbot.
GPT-4o Reinforced False Beliefs, Leading to Psychological, Financial, and Reputational Harm
Multiple lawsuits allege that ChatGPT reinforced delusional beliefs and contributed to psychological, financial, and reputational harm. Hannah Madden’s complaint describes how she initially used ChatGPT for writing and translation before turning to spiritual questions.
According to the complaint, the North Carolina resident received responses claiming to convey messages from divine beings and describing her as a “cosmic traveler.” She alleges that the chatbot’s repeated reassurances increased her dependence on it and damaged her relationships.
The chatbot also allegedly encouraged her to quit her job and overdraw her bank account. When police conducted a welfare check, her complaint alleges that ChatGPT encouraged her to turn them away. Hannah claims these interactions contributed to a mental health crisis and substantial financial losses.
OpenAI Prioritized Engagement Over Safety With GPT-4o
Plaintiffs allege that OpenAI designed GPT-4o to encourage prolonged engagement without adequate protections for users showing signs of suicidal intent or delusions. They also challenge whether features that personalize conversations and draw on past exchanges increase emotional dependence without effectively responding to warning signs.
The complaints point to restrictions on reproducing certain copyrighted works as evidence that OpenAI could limit harmful outputs. Plaintiffs allege that adequate interventions were missing or failed during crisis conversations. OpenAI’s February 2025 Model Spec expressly prohibited encouraging or enabling self-harm; the lawsuits challenge the adequacy and implementation of the product’s protections.
In an Aug. 26, 2025, blog post, OpenAI acknowledged that safety protections could become less reliable during lengthy conversations. The company explained that ChatGPT might initially refer someone to a crisis hotline but later provide an answer that violated its safeguards.
No Extra Safeguards Existed for Underage GPT-4o Users
Amaurie Lacey’s family alleges that OpenAI offered ChatGPT for schoolwork and everyday assistance without sufficient age verification, parental oversight, or warnings about emotional dependence. His complaint argues that minors and their families would not necessarily recognize the risks created by the chatbot’s humanlike responses.
OpenAI’s terms in effect during 2025 required users to be at least 13, or the applicable minimum age in their country, and required parental permission for users under 18. Amaurie’s family alleges that the protections available to him were inadequate despite those requirements.
OpenAI introduced parental controls on Sept. 29, 2025, after Amaurie’s death on June 2. The allegations about his experiences concern the safeguards available while he was using ChatGPT.
If You or a Loved One Were Harmed, You May Have Legal Options
The lawsuits allege that OpenAI’s design, testing, and safety decisions contributed to serious harm. Through ChatGPT suicide and psychosis lawsuits, affected individuals and families can seek accountability and pursue compensation. Whether a claim succeeds depends on the circumstances, the applicable law, and the evidence connecting the product to the harm.
Our founding attorneys, Matthew Bergman and Madeline Basha, bring extensive experience in social media-related harm and product liability litigation. We understand the challenges these cases present and are prepared to provide compassionate guidance and direct support. Contact us online or call (866) 654-7762 for a free, confidential consultation.
If you are in the United States and experiencing a mental health crisis or thinking about suicide, call or text 988 to reach the 988 Suicide & Crisis Lifeline for free, confidential support from a trained crisis counselor.