OpenAI’s opaque research releases frustrate the scientific community

The company’s latest approach involves utilizing external “shepherds” to determine the release protocols for three significant findings. These results address long-standing, complex challenges in theoretical fields, with OpenAI applying a distinct strategy to each. For the mathematics breakthrough, which purportedly resolves the Kakeya conjecture regarding the rotation of objects in multi-dimensional space, the company published the raw AI-generated output without offering a formal explanation. Conversely, the computer science finding will be parsed by an assembled group of external researchers, while the physics results remain under lock and key, with OpenAI deferring to a yet-to-be-named panel on if and how to share the data.

This fragmented methodology follows a turbulent period for the organization, which faced severe criticism after flooding academia with hundreds of unvetted preprints last month. While OpenAI presents these moves as a course correction designed to manage sensitive information, many in the scientific community view the selective disclosure as an obstacle to peer verification. By controlling the flow of information and delegating the burden of analysis to external groups, the firm continues to prioritize its own release narrative over the established norms of open scientific inquiry.