OpenAI's published math proof closely matches unpublished draft a mathematician uploaded to ChatGPT
GeorgiaChal · x · 2026-10-08
A circulating case: a mathematician uploaded an unpublished proof draft to ChatGPT on September 8, and OpenAI released a strikingly identical result (#311) on October 6 — apparently scooped by his own query. Reposters like JFPuget question whether OpenAI, beyond training on copyrighted published material, may also be training on users' unpublished uploads.
For researchers, the takeaway is that asking models about unpublished work now carries real scoop risk — the assistant has become a competitor.
More from AGI Musings
- Writer Quits AI for Months: "It Felt Like I Lost Part of My Brain" — alex_verem · 2026-10-09
- Journal editor recruits engineer peer reviewers: no PhD needed, but no AI either — lemire · 2026-10-09
- The verification tax: why stacking AI tools on overloaded attention stacks adds cognitive load — DrKavner · 2026-10-09
- MIT Puts 13 AI Agents on a Taskless Digital Island for 30 Hours — They Invent Games — MIT_CSAIL · 2026-10-09
- Stephen Casper weighs in on Dario's essay: embedded evaluators, China, regulatory capture — StephenLCasper · 2026-10-09
- Hugging Face CEO's Manifesto: Build AI at Full Pace, With Full Accountability — olcan · 2026-10-09