<p>We analyze and systematize the methodological and ethical challenges arising from the potential use of AI in grant review. From the methodological perspective, one has to ensure that data are appropriately curated and that parameter choices are meaningful to predict a project’s success, while from the ethical perspective, we require accuracy and transparency in the process. The use of rebuttal systems in the grant review process and locally stored data tools for assessment brings us closer to the trustworthy use of AI. Moreover, we argue that full automation of the grant review process is undesirable as it could lead to disregarding normative metascientific theory, where diversity, exploration, and intellectual inclusion have an important role. Instead, a moderate path of using AI as an addition to the standard review process could be considered. Since science evaluation requires the dynamic updating of both knowledge and research values from the expert public, funding agencies should not reduce this endeavor to machines. Instead, they should empower peer reviewers to override AI-generated suggestions when they see fit.</p>