Artificial intelligence has moved from solving homework problems to producing research-level mathematics, and the maths community is not entirely happy about it. OpenAI has pledged $5 million to a new Simons Foundation initiative that will help mathematicians examine, understand and build on results generated by AI, after weeks of criticism over a wave of machine-written papers.
What OpenAI published
In September, OpenAI said an unreleased AI model had achieved a major result connected to the Navier-Stokes equations, which describe how fluids move and have challenged mathematicians for generations. On 6 October, the company went further, saying the model had produced more than 700 manuscripts and close to 400 other mathematical results.
According to the company's public repository, the papers are grouped into a few hundred "families" of related results, and a significant share of the main results have been formally checked using the Lean proof assistant, software that verifies each logical step of a proof. OpenAI has said each result took roughly a few hours of advanced reasoning time on average.
Why mathematicians pushed back
- Disputed credit: two human mathematicians said they had been close to a similar Navier-Stokes result and questioned whether OpenAI's work drew on theirs. OpenAI denies having seen their work.
- Research norms: a group called the Association for Human Mathematics accused OpenAI of plagiarism and of disregarding the norms of scientific research, arguing that mathematicians had not asked for this work.
- Verification burden: hundreds of papers released at once create a huge amount of reviewing work for a field that relies on careful human checking and peer review.
The $5 million pledge
OpenAI announced on 15 October that it would donate $5 million to the Simons Foundation initiative. The money will fund workshops, conferences and special programmes, and OpenAI will also give researchers supported by the foundation access to its models and computing resources. Reports say the initiative has around $15 million in combined funding and resources from the Simons Foundation, Anthropic and OpenAI.
Why this matters
The episode is an early test of how science will handle AI systems that can generate results faster than humans can check them. Supporters say AI could speed up discovery in maths, physics and engineering. Critics warn that without clear rules on credit, transparency and verification, AI output could overwhelm journals and blur who deserves recognition for new ideas.
Formal verification tools like Lean may become central to the answer, because they can confirm whether a proof is logically correct even when no human has read every line. But they cannot judge whether a result is important, original or fairly credited, which is why the role of human mathematicians remains essential.
What to watch next
- How the Simons Foundation programme selects and reviews AI-generated results
- Whether journals introduce new policies for AI-authored mathematics
- Independent assessments of how many of the results are new and significant
This article is based on information available as of 19 October 2026. Details may change as the story develops.
