Towards an AI policy framework in scholarly publishing

Generative artificial intelligence (AI) tools are rapidly transforming scholarly research. To harness their benefits while minimizing potential pitfalls, robust policies are urgently needed. In analyzing AI policies for authors and reviewers from major publishing organizations and leading journal groups, we found not only marked inconsistencies in permitted usage and reporting requirements, but also a general lack of clarity or guidance on ethical principles, transparency, and reproducibility. Furthermore, the restrictions set by these policies are largely unenforceable. To establish ethical norms and best practices, we propose a pragmatic, enabling principle that encourages authors to produce their best work with AI assistance, emphasizing responsible disclosure over unenforceable restrictions. Building on this principle and the strengths of existing policies, we offer ready-to-use author and reviewer policy templates. When AI systems are used beyond typical editing services, authors should report details of the tool, its usage, generated content, and prompts. Reviewer policies should articulate confidentiality risks. Our analysis and policy templates offer a roadmap for building consensus and best practices for the responsible integration of AI into scholarly routines, thereby accelerating discovery without compromising academic integrity.