OpenAI Retracts Three Math Preprints One Day After Massive Release; Informa to Sell Taylor & Francis

Key points

• OpenAI released 722 AI-generated mathematical preprints on October 6, 2026, claiming the move would aid scientific research.
• The release included solutions to over 300 unsolved math problems, including a variation of the Riemann hypothesis, but sparked a verification crisis.
• OpenAI withdrew three of the preprints one day after the initial release due to identified errors.
• Taylor & Francis is preparing to leave Informa, a move described as an 'open secret' for years among industry observers.
• arXiv implemented a new preprint submission cap, which has generated significant debate within the academic community.
• The Institute of Physics retracted nearly 100 conference papers for containing 'tortured phrases' and poor quality writing.

Background

This week’s developments follow a period of intense scrutiny regarding AI in academic publishing. In early October 2026, OpenAI’s mass release of 722 mathematical proofs triggered a 'verification crisis' and community backlash, with critics labeling many proofs as low-quality 'slop' and raising concerns about the reliability of AI-generated research. The current retractions and the broader discussion around preprint caps and publisher sales reflect the ongoing tension between rapid AI output and traditional academic standards of rigor and transparency.

Why it matters

The rapid retraction of OpenAI’s preprints highlights the fragility of AI-generated scientific claims and the urgent need for robust verification processes. The departure of Taylor & Francis from Informa signals a major shift in the publishing landscape, potentially affecting access and distribution models. Meanwhile, arXiv’s new caps and the retraction of low-quality conference papers indicate a growing institutional effort to manage the influx of AI-generated content and maintain scholarly standards.

What to watch

Researchers and institutions will likely continue to debate the efficacy of preprint caps and the quality of AI-generated proofs. The transition of Taylor & Francis away from Informa may lead to new operational structures or partnerships. Academic bodies may further integrate retraction data into ranking systems, as seen in India’s delayed rankings, to enforce stricter compliance with ethical standards.