Scientific review and the ethical challenge of artificial intelligence

Abstract

The use of AI detectors completely undermines transparency, fairness, trust, and merit-based evaluation. These systems generate false positives that stigmatize legitimate texts and threaten the autonomy of authors whose methodological rigor is overshadowed by unfounded suspicions. Their opaque algorithms impede understanding of how they identify "AI texts," violating due process and denying appeals. Free tools are known to barely exceed 48% accuracy. Furthermore, their Western-English training discriminates against non-native research styles and excludes diverse languages. The privatization of strict platform evaluation introduces conflicts of interest and community privacy risks. Replacing human critical dialogue with automated metrics dehumanizes peer review. Instead of technological surveillance, journals should promote validated plagiarism detectors, open reviews, and require the disclosure of AI in writing, relying on expert judgment.