Researchers Criticize OpenAI Gatekeeping After New Math and Science Advances
OpenAI announced major breakthroughs in math, physics, and computer science, but researchers are criticizing its selective expert release strategy.
OpenAI has announced a series of major theoretical achievements across mathematics, theoretical computer science, and quantum physics. However, the company's decision to control how these discoveries are published has drawn sharp criticism from the academic community. Following an October incident where the AI lab unceremoniously released hundreds of unreviewed preprints, OpenAI is now using external "shepherds"—handpicked groups of domain experts—to guide the disclosure of its latest results. Despite the controlled approach, many researchers express frustration over what they view as unnecessary gatekeeping by a private corporate entity.
Major achievements across three fields
The three research outcomes target long-standing unsolved problems in their respective disciplines. In mathematics, OpenAI claims to have resolved the Kakeya conjecture across all dimensions, a problem that examines how little space is required to rotate a needle in every direction. Earlier this year, New York University mathematician Hong Wang received a Fields Medal in part for solving the three-dimensional version of the conjecture. Brown University mathematics professor Brendan Hassett noted that OpenAI's claim "looks well supported," though he cautioned that the underlying computational proof in the Lean formalization language still requires full verification.
In computer science, OpenAI claims to have disproved the exponential time hypothesis, which posits that certain complex computational problems cannot be solved significantly faster than brute-force methods. This finding connects directly to P versus NP, a famous Millennium Prize problem.
Meanwhile, the physics result investigates quantum chromodynamics within the Standard Model, examining the strong nuclear force and its relationship to Yang-Mills theory—another area tied to a Millennium Prize problem, though OpenAI's work does not solve the related "mass gap." Unlike the math result, which saw AI-generated output released publicly for open interpretation, the physics output remains entirely unreleased while OpenAI defers to its chosen expert panel.
Academic backlash against selective gatekeeping
While scientists generally find the technical results impressive, OpenAI's release strategy has sparked intense debate. Many academics argue that selecting preferred experts creates unequal access and risks granting undue authority to corporate-chosen review panels. Constantin Kogler, a researcher at the Institute for Advanced Study, welcomed the smaller, three-item announcement compared to October's flood of preprints, but criticized restricting initial visibility. Kogler stated that "it doesn't make any sense that first only a few distinguished scientists can look at them."
University of Texas at Austin physics professor Andreas Karch called the handpicking practice "a disaster," asserting that major AI labs should not act as gatekeepers who select winners and losers. Karch advocated for immediate public releases alongside expert reviews to democratize access. Furthermore, the Advisory Group on Mathematics and Artificial Intelligence (AGMAI) warned against selective access, urging OpenAI to publish its natural-language and Lean formalizations in physics without delay so that evaluation can happen in the open.
Addressing the controversy, Stanford mathematician Jared Duker Lichtman observed that finding an announcement strategy that satisfies everyone may be impossible, making it a matter of selecting the least offensive option. Georgia Institute of Technology mathematics professor Xiaoyu He added that a deeper challenge exists beyond public relations: fundamental scientific problems are now being solved rapidly without direct human contribution, yielding papers that remain inherently difficult for humans to read and verify.
What it means for developers
For software engineers and AI practitioners, these developments highlight both the remarkable power and the practical challenges of state-of-the-art foundation models. As AI systems increasingly generate complex mathematical proofs and theoretical scientific solutions, developers face growing requirements for automated verification frameworks, formal proof checkers like Lean, and robust evaluation methods.
Access to advanced model outputs remains a central requirement for teams building analytical, scientific, or reasoning applications. Developers looking to test or integrate top AI models can access them cheaply through one API key at https://apixoai.online, which provides pay-per-token access to Claude, GPT, Gemini, Grok, DeepSeek, and other leading architectures.
Ultimately, the ongoing debate between OpenAI and the scientific community underscores a broader reality for developers: as AI model capabilities advance rapidly into complex domain logic, the tools for auditing, verifying, and democratizing access to model outputs will be just as critical as the underlying algorithms themselves.
Source: Researchers are getting tired of OpenAI gatekeeping its breakthroughs — The Verge AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
