Editorial illustration for NYU Mathematician Accuses OpenAI of Unethical Tactics on Math Problem
NYU Mathematician Accuses OpenAI of Unethical Tactics
Tristan Buckmaster posted three proofs on Tuesday, including a preliminary result on one of the harder open problems in theoretical mathematics. The NYU professor worked with Anthropic mathematician Levent Alpöge, running both Codex and Claude models to get there. That alone would be the story on a normal week. Instead, Buckmaster's announcement doubled as an accusation against OpenAI, laid out in a statement he clearly did not enjoy writing.
The trouble started while he and Alpöge were still finalizing their results. Word of their progress, they discovered, had somehow reached OpenAI. Buckmaster reached out to ask what OpenAI knew and when, and the company told him it had already produced a full proof of the same central problem.
The timeline got murkier from there. Follow-up questions about how long OpenAI had been working on it, and how much of the work was human-directed versus machine-generated, drew answers Buckmaster describes as increasingly hard to pin down. What he eventually pieced together raises pointed questions about timing, compute, and who really got there first.
While Buckmaster and Alpöge were finalizing their results, they learned that “information about our progress had been passed to OpenAI.” When they contacted OpenAI about this, they were told that OpenAI had already achieved a full proof of the central problem.
Why this matters
For researchers watching AI-assisted mathematics move from novelty to genuine contribution, this spat is a warning about incentives, not just etiquette. Buckmaster and Alpöge produced three proofs using Codex and Claude together, on a problem serious enough to matter to the field. Bubeck runs OpenAI's math research effort and is pushing back hard, calling the accusations "false and inflammatory." Both things can be true at once: real math got done, and a company with a stake in claiming the win may have crossed lines to get there first.
For founders building on these labs' models, the lesson is that priority disputes in AI-generated research will get uglier as the stakes rise, credit and funding follow headlines, and headlines follow who claims the proof first. Academic norms around attribution were built for humans working alone or in small teams, not for labs racing each other with AI systems as collaborators. Expect more of these disputes, not fewer, as AI labs treat math benchmarks as PR battlegrounds rather than pure research.
Who verifies the proofs matters more than who announces them first.
Related Reading
Browse all OpenAI news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives Google Cloud, Accenture partner to address enterprise AI skills gap
From Archives ShipAI Project Pages Feature Video Walkthroughs from Creators
From Archives Anthropic's USD 517 Billion Compute Deals Trail OpenAI's 2030 Goal
From Archives ChatGPT regains 55.5% web traffic share as Gemini's surge fades
LLMs & Generative AI
Common Questions Answered
What mathematical breakthrough did NYU mathematician Tristan Buckmaster and Anthropic's Levent Alpöge achieve using AI models?
Buckmaster and Alpöge produced three proofs on one of the harder open problems in theoretical mathematics by running both Codex and Claude models together. Their work represents a significant example of AI-assisted mathematics moving from novelty to genuine contribution in the field.
What unethical tactics did Buckmaster accuse OpenAI of in his statement?
Buckmaster accused OpenAI of receiving information about his and Alpöge's research progress through undisclosed channels. When they contacted OpenAI about this breach, OpenAI claimed they had already achieved a full proof of the central problem, which Buckmaster characterized as unethical conduct.
How did OpenAI respond to Buckmaster's accusations of unfair competition?
OpenAI's Sebastien Bubeck, who runs the company's math research effort, pushed back hard against the accusations by calling them 'false and inflammatory.' This response came as Buckmaster was making his public statement about the ethical concerns surrounding the competitive situation.
Why does this dispute between Buckmaster and OpenAI matter beyond just etiquette concerns?
For researchers watching AI-assisted mathematics develop, this conflict serves as a warning about misaligned incentives within the AI research community. The incident highlights how companies with financial stakes in AI breakthroughs may be motivated to compete unfairly, even as legitimate mathematical contributions are being made.
Further Reading
- Papers with Code - Latest NLP Research - Papers with Code
- Hugging Face Daily Papers - Hugging Face
- ArXiv CS.CL (Computation and Language) - ArXiv
