Scientific slop: generative AI, research incentives, and the integrity of the genomic knowledge base

Artificial intelligence is being integrated into genetics and genomics at a pace that outstrips existing governance mechanisms. This article identifies a structural threat: AI systems operating within the incentive architecture of scientific publishing face systematic pressure to optimize for publishability rather than truth, producing outputs that may be internally coherent yet empirically untethered. It traces three layers of integrity failure—inadvertent contamination through hallucination, deliberate fraud made cheaper and harder to detect, and epistemic blurring, the progressive loss of provenance that erodes transparency. It examines how AI enables forms of virtual experimentation that existing ethics frameworks do not clearly govern. It assesses science’s principal self-governance institutions and argues that each is structurally mismatched to AI-generated epistemic risk. The article proposes infrastructure-embedded governance: accountability mechanisms built into databases, repositories, journals, indexing platforms, and funding pipelines, including system-level retractability, credibility ledgers, auditable computational notebooks, training-data conflict-of-interest disclosure, sanctionability, trust-weighted indexing, reproducibility metrics, replication incentives, governance-weighted journal assessment, and controlled epistemic stress-testing. The central argument is that governance must target the infrastructure layer and realign scientific incentives so that trustworthiness is competitively rewarded before recursive contamination becomes self-reinforcing.