Comprehensive Consideration of Ethics in AI-assisted Scientific Writing and Peer Review.

Abstract

The emergence of large language models and generative artificial intelligence (AI) is driving fundamental transformations in the ecosystem of scholarly publishing and peer review. As manuscript production enters an era of sophisticated technological assistance, it has become imperative to transition from traditional approaches focused on misconduct prevention toward a more proactive ethical framework. We propose a new standard centered on transparency, accountability, and confidentiality, presenting a clear solution to challenges associated with integrating AI into academic discourse. Regarding authorship, AI cannot be credited as an author; the final accountability for academic integrity lies only with human authors. Transparency is maintained through a tiered disclosure framework that mandates reporting based on the extent of artificial intelligence utilization. In the context of peer review, while the potential of artificial intelligence to optimize efficiency is recognized, its application must be restricted to a closed security system to safeguard against data breaches. Furthermore, this review highlights new risk factors such as algorithmic sycophancy and prompt injection attacks, emphasizing that a final verification through human expertise is essential to ensuring the integrity of the peer review process. In conclusion, we present a comprehensive regulatory revision roadmap integrating the authors' obligations for transparent information disclosures, reviewers' commitment to confidentiality and security, and editors' ethical oversight. This framework does not regard AI as an object of absolute prohibition but rather positions it as an advanced scholarly aid rooted in human intellectual accountability. The perspectives in this Special Issue will provide a practical framework to maintain academic rigor and enhance institutional trust in the era of AI.