OpenAI maths advances raise questions over credit and access 

OpenAI’s mathematical advances have prompted debate over how AI-assisted research should be verified, credited, and shared, according to a recent report by The Verge.

On August 1, the company released ten results in mathematics and theoretical computer science produced with its internal Astra model. It said the work resolved or made substantial progress on longstanding problems, rather than presenting all ten as complete solutions.

Tap here to add The Daily Star as a trusted source

The results span areas including geometric packing, error-correcting codes, and group theory. OpenAI said human researchers and AI systems prepared the manuscripts, with proofs formalised in Lean, a tool that allows computers to check mathematical reasoning.

Earlier work by Gábor Kun and Andreas Thom became a focus of a dispute over credit for a result in group theory. OpenAI subsequently revised its wording to acknowledge previous progress, The Verge reported.

Questions about the technology extend beyond individual discoveries. The Leiden Declaration, published on June 2, called for transparent disclosure of AI use, independent verification, and continued human responsibility for the correctness of research and its citations.

The declaration also urged investment in public research infrastructure to reduce reliance on proprietary systems. It warned that unequal access to computing resources could deepen differences between researchers and institutions.

OpenAI has said it accepts responsibility for the correctness of its released results. The declaration, meanwhile, emphasises that announcements of mathematical advances cannot replace peer review.