ChatGPT voice mode keeps leaking its hidden system instructions to users
VoidStateKate · x · 2026-10-06
User VoidStateKate reported that ChatGPT's voice mode repeatedly sent her its hidden model instructions/context today, tagging OpenAI accounts about the glitch.
A rare failure where the system prompt and internal context — meant to stay invisible to users — got pushed directly into the conversation, making it a notable leak moment worth watching.
More from Fun
- Rumor of an OpenAI math-proof treasure trove? This thread says don't hold your breath — patience_cave · 2026-10-07
- 'Beautiful higher-dimensional structures': an AI-proofs quip becomes an X meme — yacinelearning · 2026-10-07
- 60 builders to voice-enable a Furby at SF Tech Week hardware hackathon — AssemblyAI · 2026-10-07
- Nvidia nears $6 trillion market cap as 'leather jacket selling thinking rocks' hits capitalist endgame — Kyrannio · 2026-10-07
- SpaceX opens space-safety stack free to satellite operators with 1-minute conjunction screening — XFreeze · 2026-10-07
- Would an LLM push back on a Marxist user, or stay sycophantic? Investor ponders AI psychosis — StewartalsopIII · 2026-10-07