A mathematician asks if his ChatGPT chats trained OpenAI's prover.
German mathematician Andreas Thom spent months discussing an unsolved problem with ChatGPT itself. OpenAI later said its Astra model solved that problem, leaning on Thom's own published work. An OpenAI researcher told him his chats were not used, and Thom calls that answer misleading.
Why it mattersWhat you type into a chatbot can leave your account and shape the product. If unpublished ideas travel that way, private work becomes someone else's announcement.
▶ Related video: OpenAI announces that "non-sofic groups exist"
Read in the app — free, in 9 languages
Related stories
A Fields medallist opened an institute to do the maths of AI safety.2026-09-10A mathematician says OpenAI asked him to erase a co-author.
2026-09-10A phone survey flagged three in four suicide attempts a week early.
2026-09-10Terence Tao warns AI is mining open maths problems dry.
2026-09-10Singapore will fund four years of AI research for new PhD holders.
2026-09-09