OpenAI announced three AI-generated results across mathematics, computer science, and physics, but researchers questioned its selective release process.
OpenAI’s October 26, 2026, announcement of three AI-generated research results prompted debate among scientists about how the company is making the work available. The findings concern mathematics, theoretical computer science, and physics, but each is being shared differently, leaving researchers with uneven access to the material.
OpenAI released its mathematical output, which the company says resolves the Kakeya conjecture in all dimensions. Brown University professor Brendan Hassett told The Verge the claim appeared well supported, while noting that he had not yet examined its Lean formalization. The computer science result, which claims to disprove the exponential time hypothesis, is set to be explained with help from an external group of researchers. OpenAI has not released the physics materials; it says it will consult outside experts on how to share them.
The physics work concerns quantum chromodynamics and its connection to Yang-Mills theory. The article notes that the result does not address the unresolved mass-gap problem. These descriptions reflect OpenAI’s claims, not independently established conclusions: the findings still require scrutiny by specialists.
Several researchers interviewed by The Verge welcomed outside consultation but objected to OpenAI choosing who gets early access or delaying public release. The Advisory Group on Mathematics and Artificial Intelligence had also urged AI companies against giving selected researchers exclusive first access, citing risks of unequal access and delays. It called on OpenAI to release the natural-language and Lean outputs for the physics result, which it said OpenAI had indicated were complete.
The disagreement follows OpenAI’s release of hundreds of preprints earlier in October, which mathematicians said would take substantial time to assess. The dispute now centers not only on whether the new results hold up, but also on whether broad, timely access can help the research community verify and explain them.