OpenAI Faces Fresh Plagiarism Claim Over Mathematical Proof

OpenAI is facing a fresh allegation that it passed off a mathematical proof as its own work without crediting whoever produced it first. The claim comes from researcher Valerio Capraro, who posted the accusation on X, where it was picked up and argued over by commenters comparing the disputed result against earlier published work. OpenAI has said nothing publicly about the charge, and no full side-by-side comparison has been released.

The specifics remain thin. Capraro's post does not appear to lay out a line-by-line match between the proof in question and any prior source, which leaves the case resting on resemblance rather than demonstrated overlap. That vacuum has pushed observers to do their own forensic work, hunting through papers and problem sets for telltale similarities and disputing how much shared material is enough to constitute theft. Without a documented comparison, the accusation is an assertion rather than a finding.

This is not the first time the company has been accused of absorbing someone else's mathematical output. Earlier model releases drew similar complaints from mathematicians who noticed proofs and solutions that tracked their own unpublished or hard-to-find results. Those episodes never produced a formal resolution, and they raised the same question now in play: whether a match reflects training-data ingestion, independent rediscovery of a known theorem, or something between the two that current disclosure practices simply do not describe.

Mathematics makes these disputes sharper than most fields. A proof is a discrete, checkable artifact, and while two people can reach the same theorem independently, the intermediate steps, notation choices and ordering of lemmas tend to carry a recognizable signature. That makes plagiarism claims in math more verifiable than in prose and harder to dismiss as coincidence when those fingerprints line up. It also means the accusation can be tested rather than merely argued about.

What happens next hinges on whether Capraro or someone else publishes a detailed comparison. A clear match would put pressure on OpenAI to explain how it trains and evaluates models on mathematical benchmarks, and could force a correction or an attribution. A superficial overlap would send the episode into the same unresolved pile as its predecessors, with the burden of proof sitting on the accuser. The math community is watching for the evidence that settles which way this one goes.