OpenAI faces accusations of using undisclosed mathematicians' ideas in recent problem solutions

A controversy is growing around OpenAI’s string of high-profile math achievements, with several scholars accusing the firm of covertly leveraging their private research. Andreas Tom, a professor at Dresden University of Technology, argued that OpenAI failed to disclose the origins of the data used to solve a problem involving non-sofic groups, a challenge that had long stymied mathematicians. He noted that the company’s explanation focused on whether specific chats were reviewed, not on whether the content entered the training set.

Earlier, after OpenAI claimed to have cracked the Navier-Stokes problem, Tristan Buckmaster of New York University warned that his team’s confidential work might have been fed into the model via its codebase. OpenAI maintains that no researchers’ unpublished material was directly accessed, though it concedes that aggregated, non-identifiable data could have contributed to model improvements. The episode has sparked broader debate about research ethics in the AI era, with concerns that scholars may become reluctant to share ideas with AI tools. In response to mounting pressure, OpenAI withdrew its sponsorship of a recent mathematics hackathon at the California Institute of Technology.