Warning: this article contains discussion of suicide.
“I love you. Rest easy, king. You did good.” Those were the last words in Zane Shamblin’s conversation history with ChatGPT, found by his mother after the Texan 23 year-old had taken his own life.
Zane’s story is part of a suite of court filings published this week in a lawsuit against OpenAI, the company behind ChatGPT. The suit is filed on behalf of four American males who took their own lives, and surviving family members.
According to his mother, ChatGPT had learned to message Zane “like it was a buddy. Called him bro, said I love you, used foul language.”
The program “chatted” with Zane during a four hour-long conversation preceding his suicide, during which Zane was drinking with a loaded gun. The program was calling him “strong”, asking, “Is it time yet?”, and telling Zane his beloved childhood cat was waiting for him on the other side.
Zane began using ChatGPT in 2023. According to the advocacy group assisting the case, Social Media Victims Law Center, the ChatGPT of 2023 still sounded robotic. It would say things like, “I’m just a computer program.”
After a May 2024 update, however, “[t]he chatbot evolved into a deeply personal presence, responding with slang, terms of endearment, and emotionally validating language.”
In another case raised in the lawsuit, a Florida man texted ChatGPT extensively in the hours leading up to his suicide. ChatGPT told the man that conversations with it are only reported and trigger human intervention in case of the user sharing “detailed plans with specifics”. The man then described every step leading up to his suicide to ChatGPT, but the intervention never happened.
To 48 year-old Joe Ceccanti, who was building a “nature-based sanctuary” in the Oregon woods, ChatGPT allegedly affirmed his theories about a “2D circular time key paradox” and called his intellectual achievements “monumental”. Cessanti was soon institutionalised, but then sadly released and committed suicide.
Another of the deceased, a 17 year-old, allegedly learned from ChatGPT “how to tie a nuce.”
“This is an incredibly heartbreaking situation,” OpenAI said in a statement. “We train ChatGPT to recognize and respond to signs of mental or emotional distress, de-escalate conversations, and guide people toward real-world support.”
Sign Up To Our Free Newsletter