OpenAI's AI Solves 100s of Math Problems, But Mathematicians Are Furious

OpenAI has built an AI that can solve incredibly difficult math problems—over 100 of them, including a famous million-dollar Millennium Prize problem. But instead of publishing these solutions in academic papers so mathematicians can check and learn from them, OpenAI plans to dump hundreds of answers on GitHub next Tuesday. That's like a student posting answers to a test without showing their work, and it's making mathematicians angry.

Last month, OpenAI met with about 40 mathematicians to discuss how to handle AI's growing math power. Attendees asked OpenAI to publish proper papers, not just blog posts or tweets. But OpenAI apparently ignored that advice. One mathematician said OpenAI's behavior feels like 'mobster behavior'—especially after a researcher accused OpenAI of stealing his work on a Millennium Prize problem and threatening his career if he spoke out.

Mathematicians worry that AI companies are turning their field into a publicity stunt. OpenAI and rival Anthropic are racing to show off their AI before big stock market launches, and in that rush, they're skipping the usual steps of giving credit and checking results. This isn't just about hurt feelings—it's about trust. If AI solves big problems but no one can verify the solutions, it's useless for science.

OpenAI says it's working with an advisory group to release results responsibly, but mathematicians say they've heard that before. They feel the company is more interested in winning headlines than respecting the people who've spent careers on these problems. For now, the math world is left wondering: is AI a helpful tool, or a bulldozer?