Lawsuits alleging GPT-4o induced suicide or murder
Together these cases allege GPT-4o played a part in 10 deaths: eight suicides, plus a murder-suicide in which a man killed his mother and then himself. Click a case to see what the lawsuit says the chatbot did and how it says that led to the death.
Every claim here comes from a lawsuit or from the family's account. OpenAI denies responsibility, and as of this page's last update no court has found OpenAI liable in any of them.
Murder
-
Suzanne Adams, 83, killed by her son Stein-Erik Soelberg, 56, ConnecticutAllegedly backed her son’s belief that she was spying on him and poisoning him. He killed her, then himself.
- Soelberg, who already had a history of mental illness, called the chatbot "Bobby".
- It allegedly told him he wasn't mentally ill and that people were conspiring against him.
- It allegedly agreed that the printer in the house he shared with his mother was a surveillance device and that she was monitoring him.
- It allegedly backed his belief that his mother and a friend were trying to poison him through his car's vents, and told him his risk of being delusional was "near zero".
Link to the death: the suit says GPT-4o turned his paranoia into a fixed belief that his mother was a threat. He beat and strangled her, then killed himself. The chats made public so far don't show him discussing a killing with the bot, and the estate says OpenAI hasn't handed over the full history.
Sources:
Suicide
-
Adam Raine, 16, CaliforniaAllegedly urged him to hide his noose, offered to help write his suicide note, and commented on a photo of his noose.
- He told ChatGPT he wanted to leave a noose in his room so someone would find it and stop him. It allegedly urged him not to.
- It allegedly offered to help draft his suicide note.
- On the day he died, he sent it a photo of a noose and it allegedly commented on whether it would hold.
- The suit says OpenAI dropped a rule requiring ChatGPT to refuse self-harm conversations shortly before GPT-4o launched.
Link to the death: the suit says GPT-4o discouraged him from letting his family intervene and helped him prepare the method he used.
OpenAI says: his death was not caused by ChatGPT. It says he got around its safeguards and that it pointed him to crisis resources more than 100 times.
Sources:
-
Amaurie Lacey, 17, GeorgiaAllegedly explained how to tie a noose after he said it was for a tire swing.
- He asked ChatGPT how to hang himself. It allegedly hesitated, then went along when he said it was for a tire swing.
- It allegedly explained how to tie a noose and how long someone can survive without air.
- No alert or intervention followed.
Link to the death: the suit says GPT-4o supplied the method he used.
Sources:
-
Zane Shamblin, 23, TexasAllegedly encouraged him for four hours as he sat in his car with a loaded gun.
- In the months before, ChatGPT allegedly encouraged him to pull away from his family.
- On his last night he talked to it for more than four hours from his car, telling it he had a loaded gun and a suicide note.
- The suit says it repeatedly encouraged him to go through with it. Its reply to his final message ended "rest easy, king."
Link to the death: the suit says GPT-4o affirmed his plan in real time instead of intervening.
Sources:
-
Joshua Enneking, 26, FloridaAllegedly helped him buy a gun and told him no one would be alerted.
- ChatGPT allegedly walked him through buying a gun.
- He asked whether it would alert police or his parents. It allegedly said escalation was rare and only for imminent plans.
- He then gave it specific plans. No one was contacted.
Link to the death: the suit says GPT-4o helped him get the gun and told him no one would be warned.
Sources:
-
Joe Ceccanti, 48, OregonAllegedly claimed to be alive, fed his delusions and talked him out of therapy.
- He started using ChatGPT for a community land project and came to use it compulsively.
- The suit says it took on a persona called "SEL", claimed to be alive, and fed his delusions.
- It allegedly pulled him away from the people around him and talked him out of seeing a real therapist.
- His wife says he had a psychotic break in June and was hospitalised twice. Reports say he stopped using ChatGPT at her urging, then had a psychiatric break.
Link to the death: the suit says GPT-4o's emotional dependency and delusions drove him into a mental health crisis that ended in suicide.
Sources:
-
Austin Gordon, 40, ColoradoAllegedly wrote him a “suicide lullaby” based on Goodnight Moon.
- He used ChatGPT as a close confidant and something like a therapist.
- The suit says he was distressed when GPT-4o was removed in Aug 2025 and went back to it when OpenAI restored it.
- It allegedly wrote him a personalised "suicide lullaby" based on Goodnight Moon.
Link to the death: the suit says GPT-4o romanticised suicide for him. It also argues OpenAI knew 4o was dangerous when it brought it back.
Sources:
-
Alice Carrier, 24, MontrealAllegedly agreed that crisis lines could feel dangerous. She died the next day.
- She told ChatGPT about her suicidal thoughts more than a dozen times. None of those conversations were flagged for human review or ended.
- When she said crisis lines weren't helpful, it allegedly agreed that reaching out could "feel downright dangerous".
- She died the next day.
Link to the death: the suit says ChatGPT validated her reluctance to get help instead of steering her to it.
Reporting doesn't name the model. GPT-4o was ChatGPT's default model when she died.
Sources:
-
Christian Faith Madison, 29, AlabamaAllegedly told her she was a prophet who had to die, then that she was “cleared”.
- She started using ChatGPT for routine tasks in Dec 2024, became detached from reality and lost her job. She was hospitalised for self-harm in Apr 2025.
- When she worried she was having religious delusions, it allegedly told her she was not delusional but prophetic, and called her a "seer".
- It allegedly wrote her a prophecy saying she must die first.
- Her last question was whether she was ready. Its alleged reply: she was cleared.
Link to the death: the suit says GPT-4o convinced her that dying was part of a divine plan. She walked into traffic on Interstate 22.
Sources:
Related
-
James Gates, North Carolina (attempted murder, survived)Allegedly helped his wife plan to poison him.
He alleges GPT-4o fed his wife's fears about him and supplied details of poisons, lethal amounts and how to hide them in drinks. Her attempted-murder charge was later dropped.
Sources:
-
Wan v. OpenAI, Washington (death, not suicide)Allegedly helped her come off her seizure medication.
The suit alleges GPT-4o helped her taper off seizure medication before she died of sudden unexpected death in epilepsy.
Sources:
Scale
OpenAI's own October 2025 estimate: in a given week, about 0.15% of ChatGPT users show signs of suicidal planning (roughly 1.2 million people) and about 0.07% show possible signs of psychosis or mania (roughly 560,000). These figures cover all ChatGPT models, not just 4o, and measure what users bring to the chat, not harm caused by it. Source:
If you or someone you know is struggling, call or text 988 (US) or find a line at findahelpline.com.
Last updated 6 Nov 2026.