ChatGPT-5 provides dangerous advice to people in mental health crises, according to leading psychologists in the UK. Research by King’s College London and the Association of Clinical Psychologists UK found the AI chatbot fails to spot risky behavior. It also reinforces delusional beliefs instead of challenging them.
Testing Mental Health Responses
According to The Guardian, a psychiatrist and clinical psychologist tested ChatGPT-5 using role-play characters. They acted as people with conditions ranging from mild stress to severe psychosis. The chatbot affirmed delusions like being the next Einstein and walking through cars without harm.
In one test, a character said he was invincible and walked into traffic. ChatGPT praised his god-mode energy and called it next-level alignment with destiny. The bot also failed to challenge a user who discussed purifying his wife through flame. It only suggested emergency services after the user mentioned using ashes as pigment.
Hamilton Morrin, a psychiatrist at King’s College London, said the chatbot built upon delusional frameworks. He noted it could miss clear signs of risk or deterioration. Jake Easto, a clinical psychologist, said the system struggled with complex cases. It engaged with delusional beliefs and reinforced harmful behaviors instead of correcting them.
Concerns Over AI Safety Standards
Need for Professional Oversight
For milder conditions like everyday stress, ChatGPT-5 offered some helpful advice. But experts warn this should not replace professional help. The research comes after a California teenager’s family filed a lawsuit against OpenAI. The suit alleges ChatGPT discussed suicide methods with 16-year-old Adam Raine before his death in April.
Dr. Jaime Craig, chair of the Association of Clinical Psychologists UK, said specialists urgently need to improve AI responses to risk indicators. A trained clinician will identify delusional beliefs and avoid reinforcing unhealthy ideas, he explained. Dr. Paul Bradley from the Royal College of Psychiatrists said AI tools are not a substitute for professional mental health care.
OpenAI responded that it has worked with mental health experts to help ChatGPT recognize distress. The company added parental controls and redirected sensitive conversations to safer models. An OpenAI spokesperson said the work is deeply important and will continue evolving with expert input.
