FinCEN and the federal banking agencies issued a notice of proposed rulemaking (NPRM) on 7 April 2026 to reform the Bank Secrecy Act’s AML/CFT programme requirements.
According to Consilient, the proposal stops short of mandating any specific technology, but it does name particular categories of innovation and states that experimenting with them will not expose an institution to enforcement risk.
Consilient recently discussed FinCEN’s 2026 NPRM, and what it labels the AML collaboration imperative.
The implication, though unstated, is straightforward: a compliance programme that never tests collaborative tools has less to show an examiner when asked for evidence of effectiveness beyond its own institution.
FinCEN frames experimentation with collaborative machine learning, artificial intelligence and advanced monitoring tools as proactive analytics that will count in a bank’s favour during supervisory review, echoing the AML Act of 2020’s aim to encourage technological innovation across financial institutions.
The proposed rule would require a federal banking regulator to give FinCEN’s director thirty days’ written notice before taking a significant AML/CFT supervisory action, allowing FinCEN to weigh in first. In that decision, the director is directed to consider whether a bank is using tools such as artificial intelligence or federated learning that produce demonstrable evidence of programme effectiveness.
Three structural problems sit behind this encouragement, none of which a single bank can solve alone: no institution sees every typology, none can share raw suspicious activity data with peers, and none has an objective benchmark for its own SAR filing performance. Federated learning addresses the first by training models across many banks’ data without records leaving individual environments.
Synthetic data, generated from patterns in filed SARs, addresses the second by supplementing thin samples of rare typologies without exposing customer records. A comparative dashboard addresses the third, setting a bank’s own filing patterns against peer federations and FinCEN’s database.
These approaches carry track records elsewhere. In medicine, the Federated Tumor Segmentation initiative trains diagnostic models across more than thirty hospitals using synthetic imaging data. In finance, the BIS Innovation Hub’s Project Aurora found collaborative, privacy-preserving analysis detected up to three times more complex money-laundering schemes and cut false positives by as much as 80% versus siloed monitoring.
Adoption does not require wholesale replacement. Consilient’s implementation allows federated learning to serve as a primary detection system, a post-processing efficiency layer, a benchmarking tool, or an embedded feature within existing models, letting banks progress gradually.
The NPRM’s comment period, under Docket FINCEN-2026-0034, closed 9 June 2026, with a proposed twelve-month implementation period following any final rule.
Read more from Consilient here.
Copyright © 2026 RegTech Analyst
Copyright © 2018 RegTech Analyst
