Perspective: is AI about to erode science’s legitimacy?

With artificial intelligence accelerating scientific discovery and automating complex research tasks, the rapid integration of AI in research is also creating a collective action problem that pits individual academic productivity against the long-term reliability of the scientific record. This introduces deeply destabilizing vulnerabilities into contemporary and future scholarship, characterized by fabricated references infiltrating reputable peer-reviewed venues and the imminent threat of model collapse as flawed, AI-generated literature pollutes tomorrow’s model-training datasets. These vulnerabilities threaten to permanently compromise the credibility of science as the most dependable knowledge engine we have, if the scientific community does not address them now through proper accountability and systemic governance frameworks.