OpenAI Creates a Math Advisory Group as It Reports More Than 100 Solved Problems

OpenAI Creates a Math Advisory Group as It Reports More Than 100 Solved Problems

September 22, 2026

OpenAI has formed a mathematics advisory group after reporting that its AI systems have resolved more than 100 open problems, TechCrunch reports. The announcement sits at the intersection of two questions: can AI make useful contributions to research, and how should labs communicate claims when outside experts have not yet checked every result?

A mathematical problem is not considered solved because a model produces a plausible proof. The argument must be logically valid, use definitions correctly, and withstand scrutiny by specialists. Some claims may concern problems that are open in a particular formulation but have known variants or additional assumptions. Careful documentation matters.

Why an advisory group matters

Outside mathematicians can help evaluate whether problems are genuinely open, whether the solutions are novel, and whether the proof methods are sound. They can also advise how to share results with the research community. OpenAI’s announcement says the group is not empowered to slow or redirect the lab’s research, so its role should be understood as advisory rather than an independent approval authority.

The strongest process would publish complete proofs or formal artifacts, identify which parts used AI, and allow independent experts to reproduce the reasoning. If work is shared only as a headline count, researchers cannot judge the depth or reliability of the contribution.

What “more than 100” could mean

The number of solved problems is a rough measure, not a measure of importance. Problems vary widely in difficulty, and a system could generate many solutions to narrow or low-impact questions while struggling with a smaller number of deep conjectures. It is useful to know how problems were selected, whether they were drawn from public benchmarks, and how many attempts failed.

AI may still be valuable even when its role is assistive. A model can search for patterns, formalize an argument, generate candidate lemmas, or help a human explore a proof. These roles should be described precisely so that credit and limitations are clear.

What researchers should watch

The advisory group’s membership, charter, review procedures, and publication record will show whether it provides meaningful scrutiny. Independent papers and formal verification results would help separate discovery from promotional claims.

AI-assisted mathematics may become a practical research tool, but the standards of proof do not change. OpenAI’s new group is a step toward outside engagement. The next test is whether the lab’s claims become reproducible contributions that mathematicians can examine in detail.

Sources

• TechCrunch: OpenAI forms a math advisory group
• Lean theorem prover
• American Mathematical Society