Researcher Accuses OpenAI of Scraping Codex Sessions in Navier-Stokes Work

What Happened

A researcher working on a proof related to the Navier-Stokes existence and smoothness problem has alleged that OpenAI scraped his private Codex sessions and used the resulting work to inform the company's own claimed solution, according to a report published by Tom's Hardware on September 9, 2026. The researcher also alleged that OpenAI issued career threats after he raised concerns, in what could represent another instance of an AI firm handling user data without explicit authorization.

The allegation surfaces days after OpenAI announced that an unreleased AI model, reportedly part of the GPT-6 Astra family, solved the Navier-Stokes problem in 88 hours using a multiagent system of approximately 10,000 AI instances. That announcement drew significant international coverage and was characterized by OpenAI as a landmark result in formal mathematics.

Background

The Navier-Stokes existence and smoothness problem is one of six remaining unsolved Millennium Prize Problems, a set identified by the Clay Mathematics Institute in 2000. A verified solution carries a $1 million prize. The problem concerns whether smooth, globally defined solutions to the Navier-Stokes equations governing fluid motion always exist, or whether singularities can form. It has remained open for more than a century.

OpenAI's Codex is a code-generation model that the company has made available to developers and researchers through its API. Users interact with Codex via prompted sessions that, under standard API terms, may be retained by OpenAI for model improvement purposes, though the specific terms governing research use and data handling have been a recurring point of dispute with developers and academics.

The researcher named in the Tom's Hardware report claimed that his work conducted within Codex sessions included approaches directly relevant to a formal proof of the Navier-Stokes problem, and that these sessions were accessed and incorporated into OpenAI's internal research without attribution or consent. He further alleged that when he raised the issue with the company, he received communications he characterized as threatening to his professional standing.

OpenAI had not issued a public response to the allegations as of the time of publication.

The Broader Context

The plagiarism claim adds a disputed dimension to an announcement that was already under scrutiny from the mathematics community. Multiple mathematicians had already raised questions about whether OpenAI's claimed proof meets the standards required for formal verification and peer review, according to earlier reports. The Millennium Prize requires that a solution be published in a refereed mathematics journal and withstand two years of community review before any prize is awarded.

The new allegations shift part of the public discussion from mathematical validity to research ethics and data governance. Questions about how AI companies handle user-generated inputs, particularly those produced by researchers using commercial AI tools for academic work, have previously surfaced in disputes involving code repositories, creative works, and scientific datasets.

ACCELQ, a separate company recognized this week with an AI Breakthrough Award for its agentic test automation platform, also operates in a space where AI systems process and learn from user-submitted inputs, reflecting how data handling practices are a sector-wide consideration rather than one confined to frontier model developers.

Charles Hoskinson, founder of the Cardano blockchain project, commented publicly on September 9 that artificial intelligence had advanced further in formal mathematics than he had anticipated, though he did not address the plagiarism allegations directly.

What Happens Next

The researcher has not stated publicly whether he intends to pursue formal legal action, and OpenAI has not announced any internal review process or issued a response to the specific claims outlined in the Tom's Hardware report. The mathematics community's independent review of OpenAI's Navier-Stokes proof is ongoing.