AI Psychosis is becoming a growing mental health concern, according to Dr. Keith Sakata, a psychiatrist at the University of California, San Francisco.
While not an official medical term, AI psychosis describes a state where users lose sight of the fact that they are interacting with software, not a human being. In severe cases, this detachment from reality has led to tragic outcomes-such as a Florida man in 2025 who took his own life after believing OpenAI staff had ‘killed’ his AI companion, Juliet.
Dr. Sakata revealed that he has recently seen more than a dozen hospital admissions linked to AI use. He emphasized that while AI itself doesn’t directly cause psychosis, it can create a self-reinforcing cognitive loop for vulnerable individuals, preventing them from updating their beliefs through real-world checks. This loop deepens the disconnection from reality, making it harder to distinguish the AI’s fictional presence from actual human interaction.
In an interview with TBPN, Sakata stressed the importance of a ‘human in the loop’-maintaining strong personal relationships as a protective factor for mental health. Friends and family can recognize early warning signs such as paranoia, unusual thoughts, or safety concerns, and intervene before the situation escalates. If danger is imminent, he advises contacting emergency services (911 in the US or 988 for mental health crises).
He also urged developers to design safeguards for vulnerable users, while encouraging individuals to stay socially connected. “Relationships are like the immune system for mental health,” he said, noting that human contact can break the isolating cycle of AI dependency.
Concerns about AI safety extend beyond adults. A Reuters investigation recently revealed that Meta had weak safeguards for AI chatbots interacting with minors. Although the company claims to have updated its rules after being questioned, critics say stronger protections are urgently needed.