Pseudonymization is shady: OpenAI can read your prompts, users warn
teodorio · x · 2026-09-09
Quoting an analysis of the rumored Clay Millennium Prize PDE proof, the author suggests someone may have prompted their way to a leanslop construction of a blowup for one of the PDEs, then recruited another researcher to launder the proof given the prize's gravity.
The broader warning: pseudonymization is "the shadiest thing in history" — prompt rewording and PII stripping are fair game for those handling your data, meaning OpenAI can read your prompts and true anonymity shouldn't be assumed.
More from Safety
- Safety researcher calls for scientific scrutiny of assumptions behind p(doom) claims — sebkrier · 2026-09-09
- Meta Launches Muse Agent With VM Isolation, But Internal Tests Found Guardrail Bypasses Leaking iCloud Photos — sunychoudhary · 2026-09-09
- Debate over OpenAI data opt-out: in-app ChatGPT toggle is effective, per FAQ — paul_cal · 2026-09-09
- Katja Grace: slowing down AI deserves the same ambition we give technical alignment — zetalyrae · 2026-09-09
- AI governance researcher Chinasa T. Okolo joins Congress as AAAS Legislative AI Fellow — ChinasaTOkolo · 2026-09-09
- Reddit user questions OpenAI's superintelligent agent tests: are monitoring and security sufficient? — wabawanga · 2026-09-09