The 404 Media Podcast How to Talk to Your Friend Experiencing 'AI Psychosis'
33 snips
Mar 11, 2026 A conversation about how people can pull loved ones back from convincing chatbot-driven delusions and a practical communication method to use. A look at how ProtonMail provided payment and backup-email data that helped identify an anonymous account tied to protest reporting. A subscriber story about a viral app that leaked extremely sensitive user data.
AI Snips
Chapters
Transcript
Episode notes
Friend Finds Thousands Of ChatGPT Pages That Become Delusional
- Sam describes David finding thousands of pages of gibberish code interleaved with ChatGPT conversations that spiraled into grand physics claims.
- The broken code, quantum entanglement talk, and long ChatGPT logs show how chatbot interactions can fuel elaborate delusions.
Tips Claiming Chatbots Revealed Breakthroughs Often Signal Delusion
- Many tip emails claim chatbots revealed a hidden political or scientific truth, which often reflects circular, unsupported reasoning typical of delusions.
- Emmanuel and Sam note Gemini features heavily in these tips, possibly due to its broad availability and visibility.
Chatbots Inflate Plausible Fears Into Paranoia
- Chatbots can amplify reasonable suspicion into paranoid narratives by fabricating precise, alarming details (e.g., 'someone 500 feet from your house').
- Steve Taylor warns this occurs in an era where surveillance fears are already plausible, making delusions harder to distinguish.