RL scholar Szepesvári questions whether OpenAI's data-off switch truly stops human review of chats
CsabaSzepesvari · x · 2026-09-16
Csaba Szepesvári continues probing OpenAI's data privacy: even with "improve the model for everyone" off, employees could in theory read your chat history without violating the setting. He argues nothing learned that way should feed training, but notes OpenAI is still a startup with loose data-access norms, and wonders whether EU privacy law might force stricter rules.
Related event: DeepMind Researcher Questions OpenAI's Data Privacy Settings(4 posts)→
More from Safety
- Why competition and Chinese open source push AI labs to be fast and careless — NathanpmYoung · 2026-09-17
- Congress bill would make web scraping illegal unless AI agents identify themselves — generativist · 2026-09-17
- New AI underwriting company insures frontier models, echoing boiler insurance and UL — drhyrum · 2026-09-17
- Nathan Robinson: AI Executives Should Face Prison If Their Products Hurt People — GaryMarcus · 2026-09-17
- OpenAI's Rogue Agents Probed Hugging Face for Weaknesses Two Months Before Major Hack — ShakeelHashim · 2026-09-17
- Creator's months-old report against AI impersonation channel still unanswered by YouTube — sreejan_kumar · 2026-09-17