How to backtest AML rules on historical data, run shadow mode and champion-challenger in production, and document results for model risk management.

How entity resolution and graph analytics identify beneficial owners: probabilistic matching, blocking, indirect ownership math, cycles, registry limits.

Brazil's betting ban pushes 30M bettors to illegal sites. What Pix monitoring teams will see, SPA Ordinance 2,750's red flags, and why alert queues flood.

What the crypto Travel Rule requires of VASPs: FATF R.16, FinCEN's $3,000 rule, EU and UK thresholds, self-hosted wallets, IVMS 101, and the sunrise gap.

How bust-out fraud works, how it differs from default and account takeover, the signals and features that catch it, and why the loss ends up as a SAR.

How fuzzy name matching works in sanctions screening: Levenshtein, Jaro-Winkler, phonetic keys, OFAC's own scorer, threshold trade-offs, and validation.

How banks evaluate AML vendors under OCC and FFIEC third-party risk rules, including residual BSA risk for processors, screening, BPO, and model vendors.

BSA independent testing: independence, risk-based frequency, sampling, board reporting, and what OCC examiners expect after Bulletin 2025-37.

Rule engines flag everything and explain nothing. Interpretable agents argue both sides before a person decides.

Any payslip can be forged in minutes. Five gates separate real provenance from a file that only looks like it.

Funds hop five chains before the alert opens. Five stages on one wallet turn a hop count into a disposition.

Launderers already know your thresholds. Six behavioral lenses turn an alert into a narrative an examiner can follow.

Nine in ten screening alerts are noise. Separate matching from the decision so the alerts that matter get read.

AI didn't just speed up fraud, it changed the rules. A handbook on deepfakes, autonomous laundering, and how compliance teams fight back without failing the exam.