Banks build the transaction graph — accounts as vertices, edges where money flowed in the last 90 days. A cluster of n accounts that all paid each other forms a near-clique, and that pattern is the textbook fingerprint of layering: cycling funds through shell entities to obscure origin. Real AML systems (Palantir, ComplyAdvantage, Quantexa) don't run exact maximum-clique on 100M-vertex graphs — they sample and approximate. They look for dense subgraphs and rank by anomaly score. The same algorithm shows up in bioinformatics: maximal cliques in protein-protein interaction graphs are candidate functional complexes. And in social-network feeds: tight friend-cliques are how Facebook decides whose photos you actually want to see.
09
Money laundering rings
Anti-money-laundering teams look for cliques in the transaction graph. Tightly-connected groups of accounts moving money in circles are how laundering looks from the outside.