In crisis?Call or text 988 · Call 911 for immediate danger

See all crisis resources →
All news

Tech & Culture · United States

Scientific American: AI therapy bots have benefits, real risks, and more risks

A careful look at what the evidence actually shows about chatbot mental-health interventions.

These are short summaries of research and reporting published elsewhere. We link to the original. Mental Hum does not report these stories.

Mental Hum is owned and operated by Dean Shaban

Scientific AmericanNew York202411 min read

Scientific American's feature on AI therapy chatbots is one of the more nuanced popular treatments of the topic. It distinguishes between purpose-built clinical tools that have actually been studied (Woebot, Wysa) and general-purpose LLM companions that have not.

The piece is unusually direct about the failure modes: hallucinated clinical claims, scope drift into self-harm and suicide territory, and the privacy implications of sharing intimate detail with consumer platforms. It also gives credit where due, on access, stigma, and cost.

Read Scientific American's full feature for the research it covers.

Read the original at Scientific American

Mental Hum is a reading list, not a publisher. The summary above is our own editorial framing; the reporting and analysis live at the source.