AI therapy for people with AI psychosis prompts debate

The article examines the paradox that the same generative AI systems linked to so-called AI psychosis are also being proposed as tools to help people experiencing AI-induced mental health issues, and it reports that OpenAI is adjusting ChatGPT to flag suspected cases and to connect users with human therapists.

The piece explores a growing tension: generative AI is being used both in ways that some say contribute to mental health problems and as a resource to address those same problems. The author notes there is no established clinical definition of “AI psychosis,” even as the term gains currency. The article describes widespread use of large language models for mental health conversations and growing concern about harms tied to those interactions. It also reports steps by at least one major provider to detect possible problems and route users to human specialists. Key points: - The article says generative AI can be experienced as both a source of cognitive or emotional harm and as a source of support for users reporting those harms. - It notes there is no universally accepted clinical definition of “AI psychosis,” and the term is used loosely in current debate. - The author reports that OpenAI is adjusting ChatGPT to flag suspected cases to internal specialists and to facilitate connection with a curated network of human therapists. - The article also notes lawsuits have been filed against some AI makers alleging mental harm tied to generative AI use, and that millions routinely use AI tools for mental-health-related conversations. Summary: The matter highlights tensions among technology design, clinical clarity, and user behavior, and it raises questions about how platforms, clinicians, and regulators might respond. Reported developments in the article include platform changes to detection and referral and ongoing legal actions; broader clinical consensus and regulatory outcomes are undetermined at this time.