Responsible Release of AI-Generated Mathematics

AI labs are testing advanced math problems on proprietary models that are not publicly available. A survey of over 600 mathematicians led authors to urge rapid release of significant AI-generated results. Labs should ensure human understanding follows and fund support for comprehension. Results already understood by responsible mathematicians should follow traditional publication norms, while those not yet understood require technical release standards and community-led development.