Mathematicians Tristan Buckmaster and Levent Alpöge have issued a statement claiming that OpenAI's internal models may have solved unpublished mathematical problems that they themselves had been working on.
SecurityTristan BuckmasterOpenAI
Math Researchers Express Concern Over Alleged Use of Unpublished Research by OpenAI
This article is a translation. Read the Japanese original
According to a statement from Mr. Buckmaster, an OpenAI representative claimed that an internal OpenAI model had derived a solution for the finite-time blow-up of the "forced Navier-Stokes equations." This problem is reportedly remarkably similar to the one Mr. Buckmaster and his team had been researching. Mr. Buckmaster pointed out that while OpenAI responded that they "do not refer to user data," they did not provide a clear answer as to whether dialogues regarding research content are reflected in the models' training.
Furthermore, Mr. Buckmaster stated that when he received a proposal from OpenAI regarding the order of publishing research results, he was met with intimidating language such as "Are you trying to ruin your career?" He is calling for transparency, stating that this situation could damage the processes within the mathematics community.
Source:
- More questions about whether researchers can trust OpenAI with unpublished math (Hacker News Frontpage, 2026-09-10)
- Researcher's statement
Sources
- More questions about whether researchers can trust OpenAI with unpublished math （Hacker News Frontpage, 2026-09-10}）
- Researcher's statement