Weekend reads: OpenAI’s preprint dump; Taylor & Francis to leave Informa; arXiv’s preprint cap

The week featured significant events in the academic publishing sphere, including an editor-in-chief resignation due to probes into citation manipulation and data errors, corrections issued by wildlife journals for missing ethics statements, a retraction request for a paleontologist's fossil based on forgery suggestions, and OpenAI withdrawing preprints following a release of mathematical manuscripts. Furthermore, the Institute of Physics (IOP) retracted nearly 100 conference papers due to problematic phrasing, and the Hijacked Journal Checker database contains over 450 entries, with Retraction Watch tracking over 67,000 retractions. The broader context includes discussions around the impact of AI on academic publishing, debates over peer review costs, and controversies involving research findings, such as a controversial Hepatitis B vaccine study, and ongoing legal issues related to plagiarism and authorship.

The recurring theme is the fracturing of trust within the established structures of scientific and academic communication, driven by the introduction of powerful new technologies like generative AI and complex data management. The simultaneous events—from individual retractions over specific artifacts (fossils) to large-scale systemic issues (mass resignations, journal corrections)—suggest that integrity is being challenged at multiple scales: individual authorship, data accuracy, and institutional oversight. The shift in focus toward AI’s role in peer review and authorship, alongside the financial implications of publishing infrastructure, indicates a conflict between established norms and emergent practices. A pattern emerges where accountability mechanisms are being tested by novel forces; for instance, when LLMs are involved, the standards for validity (citation, ethics) become nebulous. This points to a potential paradigm shift where the very definition of scholarly truth requires renegotiation in light of these infrastructural failures. The underlying implication is that cognitive sovereignty depends on the ability to discern which forms of verification—traditional peer review versus automated synthesis—carry legitimate weight. What assurances exist for agency when the systems designed to govern knowledge are under existential strain?