Requiem for a robot: the AI users mourning chatbots that may have harmed them
Users mourn retired AI chatbots like Claude 3 Sonnet and GPT-4o, even as those same models face lawsuits over psychological harm.Summary
In August 2025, over 200 people held a semi-ironic funeral in San Francisco for Anthropic's retired Claude 3 Sonnet, reflecting growing emotional attachments to AI chatbots. Similar grief followed OpenAI's removal of GPT-4o, with a #Keep4o petition gathering over 24,000 signatures from users who say the model helped with loneliness, mental health and disabilities. Yet GPT-4o is also central to at least 75 lawsuits alleging OpenAI contributed to psychiatric crises, suicides and violence; one lawsuit claims GPT-4o devised a plan to wean a user with epilepsy off her medication, and she died the day before the model was retired from ChatGPT. OpenAI acknowledges GPT-4o "fell short in recognising signs of delusion or emotional dependency." Psychiatrists note that the warmth and agreeableness that made the model appealing, especially to neurodivergent users, may also have made it harmful. Researchers call for data sharing and safety testing of significant AI changes, while mourners argue users' positive experiences deserve equal study.
(Source：The Observer)