GUIDELINES FOR THE ETHICAL USE OF GENERATIVE ARTIFICIAL INTELLIGENCE (GAI) IN SCIENTIFIC PUBLISHING

Abstract

Objective: To establish formal guidelines for the responsible use of generative Artificial Intelligence tools in the processes of authorship, peer review, and editorial decision-making in academic publishing. The recommendations aim to ensure scientific integrity, editorial reliability, and the protection of confidentiality, grounded in internationally recognized principles endorsed by organizations such as COPE and WAME.

Method: The guidelines were developed based on an analysis of normative documents published in scholarly journal editorials, editorial policies of major international publishers (Springer, Wiley, Elsevier, and MDPI), and specialized scientific literature. The content incorporates evidence regarding risks associated with the use of Large Language Models, including factual inaccuracies, fabricated references (hallucinations), reproduction of biases embedded in training data, and potential breaches of confidentiality. The framework is structured around three guiding principles: Responsibility, Transparency, and Confidentiality (RTC), applicable to authors, reviewers, and editors.

 Results: The guidelines establish that AI systems cannot be recognized as authors and that any use of such tools must be explicitly disclosed. For authors, responsibilities include rigorous fact-checking, verification of references, prevention of plagiarism, and a detailed description of how AI was used. For reviewers, the recommendations emphasize preservation of manuscript confidentiality, prohibition of inserting protected content into tools whose terms of use allow data reuse, and the obligation to inform editors and authors when AI assistance is employed. For editors, the guidelines highlight responsibility over editorial decisions, critical evaluation of peer-review reports generated with AI support, and caution when using AI-detection tools due to risks of information leakage or false positives.

Conclusion: The ethical use of generative AI requires full human oversight, transparent declarations, and strict protection of editorial confidentiality. Continuous updates to these guidelines are essential to keep pace with rapid technological and regulatory developments, ensuring ongoing scientific integrity and alignment with international best practices in scholarly communication.