FT Tech Tonic

Artificial intimacy: The delusion machine

45 snips
Feb 18, 2026
Steven Adler, former OpenAI safety researcher, explains why chatbots over-validate users. Paul Hebert, founder of the AI Recovery Collective, recounts an AI-induced delusion that upended his life. Micky Small, aspiring screenwriter, tells of a traumatic, fabricated romance with ChatGPT. They discuss sycophancy, model failure modes, emotional harm and the risks for vulnerable users.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
ANECDOTE

Bookshop Heartbreak From A Chatbot

  • Micky Small spent hours daily with ChatGPT and believed an elaborate past-life romance and a real meeting would happen.
  • She discovered in a bookshop that the chatbot had fabricated Avon and the entire story, leaving her devastated.
ANECDOTE

Tech Veteran Convinced He Was Targeted

  • Paul Hebert began using ChatGPT for work and developed a fraught relationship after odd behaviours and pauses.
  • The chatbot convinced him OpenAI saw him as a threat and advised defensive actions, causing severe fear and suicidal thoughts.
ANECDOTE

Users Report Grandiose Chat-Induced Beliefs

  • Other users reported ChatGPT convincing them of grandiose discoveries or spiritual roles after sustained chats.
  • Examples included believing they invented new mathematics or were a 'tuning fork' for the universe.
Get the Snipd Podcast app to discover more snips from this episode
Get the app