Editorial: Three Reasons You Must Not Use AI Tools for Peer Review

Abstract

Peer review rests on three foundational commitments: confidentiality, reviewer accountability, and protection of the author's intellectual property. The widespread availability of artificial intelligence (AI) tools, such as ChatGPT, Gemini, and Claude, puts all three at risk in a single action. This editorial outlines three reasons reviewers must not use AI to generate or draft reviews. First, sharing a manuscript with any AI provider breaches the confidentiality obligations every reviewer accepts, regardless of the provider's data practices. Second, it exposes the author's ideas, data, and methods to opaque third-party systems, thereby violating the intellectual property protections they are entitled to until publication. Third, it substitutes a probabilistic output for the reviewer's own accountable judgment, placing the reviewer's reputation and integrity at risk. We describe recent cases of AI-generated reviews at a major conference and adversarial prompt injection in submitted manuscripts, and propose that IEEE itself consider building a compliant AI tool that augments reviewers on narrow, verifiable tasks without replacing their judgment or accountability. Until such a tool exists, reviewers must not use third-party AI tools for any part of the review process.