News · 2026-09-08

Buckmaster says OpenAI asked him to drop his Anthropic coauthor

Tristan Buckmaster, a mathematics professor at NYU's Courant Institute, published a signed statement on September 8, 2026 alleging that OpenAI twice proposed removing his collaborator Levent Alpoge from authorship of a paper about the Navier-Stokes problem, and that when he said he would go public he was told, "Why would you ruin your career?" OpenAI's Sebastien Bubeck responded publicly that the allegations against him are "false and inflammatory." The two accounts of the same week are difficult to reconcile.

Key facts

• Buckmaster's account is a signed PDF published at cims.nyu.edu on September 8, 2026, and announced on his Mastodon account.
• He and Levent Alpoge, a mathematician employed by Anthropic, released three results the same day: finite-time blowup with smooth forcing for the incompressible porous media, Boussinesq and 3D incompressible Euler equations.
• The disputed calls took place on Sunday, September 6; Buckmaster says he first wrote to OpenAI on September 3.
• Bubeck's reply is a public post on X; TechCrunch reported the dispute the same day.

The background is that two teams were converging on the same mathematical route at the same time. Buckmaster and Alpoge had spent about a year pushing a program originally opened by Diego Cordoba and Luis Martinez-Zoroa, using large language models heavily along the way — Anthropic's Claude, OpenAI's Codex with GPT-5.6 Sol, and more recently Astra for writeups and auditing. Progress was slow until mid-August. "On August 15th, we obtained the blow up results, with smooth forcing, for both Boussinesq and Euler," Buckmaster writes. "I can say the first LLM generated proof Levent sent me was the most horrendous I have ever read; we verified it on Lean on August 22nd."

By his account, the trouble started with gossip. On September 3, hearing a rumour that Anthropic had resolved a major open problem and having been tipped that news of their progress had reached OpenAI, Buckmaster emailed a prominent mathematician there — partly, he says, so that they would have the facts. He reproduces the email in full. OpenAI replied the same day offering compute and saying it would be useful to avoid competing. After two requests to meet sooner, the calls happened on Sunday September 6, with Bubeck joining. Alpoge was not on them.

On those calls Buckmaster says he was told an internal OpenAI model had proved finite-time blowup for the forced Navier-Stokes equations — specifically "the forcing function is smooth option c and d in Fefferman," the exact route he and Alpoge had quietly chosen. "When I heard 'forced,' it was a bright red flag," he writes, because almost nobody else was working that direction and it is not where you land in a few days from a cold start.

He also describes the story shifting during the call. He was shown a prompt and told the model had simply been handed the problem statement, and that "very little human input" was used. As colleagues fed Bubeck corrections over internal chat, Buckmaster says it emerged that an entire team had been working on it, that this was one of several attempts, that the team had warmed the model up on easier problems including Euler, that even the prompt he was shown had itself been written by prompting Codex, and that an enormous amount of compute had been spent. Asked when the first prompt was sent, he says the answer eventually agreed was: within the past few days, after information about their work reached OpenAI.

Then came the two proposals. The first was a same-week concurrent release. The second was that Buckmaster alone write a paper presenting the Navier-Stokes result, acknowledging that an OpenAI internal model had resolved it. "Sebastien twice asserted that he wanted Levent removed from authorship," Buckmaster writes, "and said it would all be simple if only it were not the case that, and it was so annoying that, Levent works at Anthropic." He declined both. When he said he would go public, the reply was "Why would you ruin your career?" and then, "If you don't want me to be nice, then I don't have to be nice." He also says Alpoge later received a text suggesting a one-on-one call, on the grounds "I don't know if Tristan is being fully rational right now."

The two published timelines do not match on a basic point. OpenAI's announcement says that after completing its work on September 6, "we reached out to them to offer a concurrent release of our result and to recognize their priority in a joint announcement." Buckmaster's statement documents him writing to OpenAI first, on September 3. OpenAI's post also says it recognizes the priority of their forced Euler work and congratulates them, and states that neither its researchers nor its agents saw any of their work before publication, adding that "while unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models." Buckmaster says he asked directly whether the model had been trained on or could access their Codex sessions — into which they had put every draft for the whole project — and that he was told the model does not look up user data, but that when he asked again about training he did not get an answer.

Buckmaster is careful about what he is not saying: "I have not seen OpenAI's proof. I do not know what their model did, or how. I do not know whether our data was used. I am not accusing anyone of anything." He also gives away the credit he could have kept, writing that in view of the body of work, "I believe Luis Martinez-Zoroa deserves a Fields Medal," and calling the moment "a Deep Blue-Kasparov moment" that the community should discuss slowly. The honest caveat here is that this is one side's contemporaneous account of private calls, Bubeck disputes it, and no recording or transcript has been published by either party.

Primary source, verified: read the paper →

Ground Truth. (2026, September 8). Buckmaster says OpenAI asked him to drop his Anthropic coauthor. Ground Truth. https://groundtruth.day/news/buckmaster-says-openai-asked-him-to-drop-his-anthropic-coauthor.html

@misc{groundtruth:buckmaster-says-openai-asked-him-to-drop-his-anthropic-coauthor,
  title  = {Buckmaster says OpenAI asked him to drop his Anthropic coauthor},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {sep},
  url    = {https://groundtruth.day/news/buckmaster-says-openai-asked-him-to-drop-his-anthropic-coauthor.html}
}

Topics: openai · anthropic · research-integrity · mathematics · credit · academia