Recommendations for developing editorial policies on the use of generative artificial intelligence in scholarly publishing

Abstract

This article aims to provide editors and scholarly publishers with an analytical and practical framework for reflecting on the design, drafting, and implementation of editorial policies governing the use of artificial intelligence (AI) in scientific publishing. Based on a review and systematization of recent international policies and guidelines, it identifies shared principles and emerging tensions related to authorship, transparency, responsibility, confidentiality, and the detection of AI use, as well as the limits of traceability and replicability when dealing with technologies based on probabilistic language models. The text is organized around three hypothetical scenarios for developing editorial policies on the use of AI: (1) prohibitive; (2) restrictive; and (3) permissive or experimental, describing for each their guiding principles, permitted and prohibited uses, disclosure and verification mechanisms, and implications for authors, reviewers, and editors. In addition, drafting examples are provided for all three scenarios, which may be used as initial templates to facilitate developing and adapting policies to different editorial contexts, disciplines, and institutional positions regarding these technologies.