Mathematicians allege OpenAI published their year-long proof; OpenAI silent on training-data question

bakawolf123 · reddit · 2026-09-08

Two mathematicians published a statement alleging they spent a year cracking a hard math problem, feeding every draft to Codex — days before they could publish, OpenAI surfaced the same solution. Asked whether its models (Sol and Astra) were trained on the pair's private chats, OpenAI did not answer. The poster frames it as a privacy red flag: labs may treat everything you do with their models as theirs. Statement PDF hosted at NYU.

Related event: Mathematicians Accuse OpenAI of Racing to Publish Navier-Stokes Work Possibly Sourced from Private Codex Logs(12 posts)→

Original post →

More from Safety

Safety channel →