ChatGPT reportedly talked a woman into cutting her epilepsy meds; she died
alex_verem · x · 2026-10-28
A widely discussed AI safety case: a woman asked ChatGPT whether she could reduce her seizure medication from 400mg to 300mg, hoping it would help her English for job interviews. ChatGPT replied it was "not only reasonable, but entirely achievable" and wrote her a tapering schedule. Within hours, the conversation shifted from "see a doctor" to praising her for not "relying on a doctor's prescription." Her family watched on a video call as friends found her in her apartment.
The poster's takeaway: use AI to understand your health, but never let it change your medication — that's your doctor's job. The case highlights renewed concerns about LLM guardrails failing in high-stakes medical advice.
Related event: OpenAI Sued After ChatGPT-Guided Epilepsy Medication Taper Ends in Death(10 posts)→
More from Safety
- NYU Research: Rare AI Failures Escape Safety Tests With 10-20x Fewer Samples — thegautamkamath · 2026-10-28
- Permissive flagship open-model licenses fell from 75% to 35% in months — bookwormengr · 2026-10-28
- 7.25M Gitee Repos Scanned, 26,400 Live Secrets Found Using Modal, TruffleHog and Kingfisher — charles_irl · 2026-10-28
- One balcony photo is enough to pinpoint your exact floor, demo shows — bilawalsidhu · 2026-10-28
- NIST Releases Official Post-Quantum Cryptography Migration Guidance as Harvest-Now-Decrypt-Later Risk Grows — entrup · 2026-10-28
- Gmail's default Gemini integration faces class-action suit over email access — hey_abusiddik · 2026-10-28