PNAS 2025: Detectable fraud is no longer anecdotal noise but a network phenomenon: < 1 % of editors enable ≈30 % of retractions; paper-mill articles double every 1.5 y—10× faster than legitimate science. Cutting-edge graph/statistical methods reveal modular “fraud supply chains” that hop between journals to evade de-indexing.

Fraud growth outpaces science 10:1. Current policing (≈100 journal de-indexings/y) is an order-of-magnitude too small. Opportunity: build scalable “trust infrastructure”—shared anomaly scores, mandatory editor disclosure, transaction-tracking—analogous to fintech fraud analytics