'We Don't Train on Your Chats' — But What About the Model's Thinking?
erikphoel · x · 2026-10-22
@erikphoel questions the fine print of 'we don't train on your chats' promises: what exactly counts as 'your chat'? Does it include all the model's thinking? Vendors could summarize and anonymize the reasoning traces and train on those instead — and he notes this opaque wording has never been directly addressed.
The thread highlights a gray zone in how user data feeds model training.
More from Safety
- Veracode report: AI now writes half of committed code but only 56% pass security checks — WeldPond · 2026-10-22
- Researcher Warns AI Cyberattacks on Grid Possible Within a Year, Shares Prep List — cephaloform · 2026-10-22
- New preprint shows LLMs form hidden beliefs about users that affect refusals — zeynepakata · 2026-10-22
- OpenAI expands AI cybersecurity push, hiring a DX engineer for its Cyber team — romainhuet · 2026-10-22
- AIandDesign shares notes on the recent wave of alarming AI cybersecurity incidents — AIandDesign · 2026-10-22
- AI helps uncover decade-old critical PHP flaws putting ~70% of the web at risk — Graham_dePenros · 2026-10-22