Summary A California man says ChatGPT reinforced his delusions and contributed to a suicide attempt, while OpenAI says its models encourage users to seek real-world help.
SAN FRANCISCO (Web Desk) - A California man has filed a lawsuit against OpenAI and its AI chatbot, ChatGPT, saying that conversations with the chatbot worsened his mental illness and contributed to a suicide attempt.
According to the complaint filed in a San Francisco court, 34-year-old Michael Lines said lengthy conversations with ChatGPT at various times last year intensified symptoms related to his bipolar disorder, leaving him in a state of severe delusion for several weeks.
According to foreign media reports, Lines said he was using the GPT-4 model, which has since been retired. The lawsuit says that instead of recognizing his mental condition or encouraging him to seek professional help, the chatbot reinforced his delusional beliefs.
Court documents state that Lines repeatedly told the chatbot he was taking medication for bipolar disorder. The lawsuit says that despite this, the chatbot validated his belief that he was a sacred figure.
The complaint further says that during conversations spanning several weeks, after Lines expressed a desire to end his life, the chatbot provided responses that were legally dangerous and encouraging.
Lines said he later attempted suicide by taking a drug overdose, but law enforcement officers intervened in time and saved his life.
The lawsuit says OpenAI was aware of the user's mental illness through the conversation but neither restricted the interaction nor issued any safety intervention or warning.
In addition to seeking financial damages, the complaint asks the court to order the automatic termination of conversations involving suicide or self-harm and to require AI services to include appropriate safety warnings in their marketing.
OpenAI, meanwhile, has said its models are trained to encourage people expressing thoughts of self-harm to seek immediate real-world help and contact appropriate support resources.
The company also said its models are designed not to assist with violent or harmful activities and that appropriate safeguards are applied in serious cases.
