84% of ChatGPT memories match what users wrote; 14% are model inferences
petesena · x · 2026-10-24
Accuracy analysis of ChatGPT memories: 84% directly match what users wrote; about 14% are conclusions the system drew itself — where things get weird.
Ask for "bands like Nirvana" and the memory becomes "User likes Nirvana" — something you never actually said.
Related event: Study Finds 84% of ChatGPT Memories Accurate, 14% Invented by the Model(2 posts)→
More from Safety
- Dev builds AI-removal-resistant watermark to fight editors like Gemini — oshekher · 2026-10-24
- For brands building AI assistants: make customer memories visible and deletable — petesena · 2026-10-24
- How to audit your ChatGPT memory tonight — plus 'Attribution Shield' fix — petesena · 2026-10-24
- User asks ChatGPT to make their job 'more interesting' — it files a fake career as memory — petesena · 2026-10-24
- Real ChatGPT memories read like therapy notes: loneliness, fear of failure — petesena · 2026-10-24
- OpenAI to showcase the next era of cyber defense on November 12 — imjustnewatai · 2026-10-24