News

U.S. regulators said Monday that financial institutions “will not necessarily” draw additional scrutiny or criticism for compliance-related weaknesses discovered while experimenting with artificial intelligence and other “innovative approaches.”

In a 3-page statement, the Financial Crimes Enforcement Network, or FinCEN, Office of the Comptroller of the Currency, or OCC, and three other federal agencies said that AI and other analytics tools may help mitigate risk and acknowledged that some institutions have developed “increasingly sophisticated” systems for spotting illicit financial flows.

The statement notes that banks that bolster their compliance programs by adopting new technology or “building or enhancing innovative internal financial intelligence units” may end up viewing past, seemingly innocuous transactions as suspicious. Others may identify gaps in their compliance programs that previously escaped notice.

But “agencies will not automatically assume that the banks’ existing processes are deficient” as a result, regulators said. “In these instances, the agencies will assess the adequacy of banks’ existing suspicious-activity monitoring processes independent of the results of the pilot program.”

Monday’s statement represents the latest instance in which federal regulators have tried to encourage new approaches towards compliance, either through technology or collaboration.

In April, during a panel at the ACAMS moneylaundering.com AML & Financial Crime Conference in Florida, senior regulators indicated an openness towards technology designed to reduce false positives, but urged institutions to run any new approach in parallel with their current systems and report newly uncovered suspicious activity.

“If you can solve that puzzle through artificial intelligence or technology in general, then I think we’ve made great strides,” Spencer Doak, the OCC’s director for Bank Secrecy Act and AML policy, said at the conference.

The latest statement is an overall positive development but may lead to closer regulatory scrutiny regardless of assurances otherwise, Jim Richards, former head of Bank Secrecy Act compliance at Wells Fargo, said in an interview Monday.

Richards, now principal of RegTech Consulting, cited the hypothetical example of a lender shunning marijuana firms, then finding out via machine learning that several clients have ties to the industry. Such a discovery would require the lender to review batches of previous transfers and possibly file dozens—if not hundreds—of suspicious activity reports retroactively, he said.

Compliance officers contacted by moneylaundering.com last year identified potentially negative reactions from regulators as one of their top reasons for not adopting AI-led systems.

Some voiced concern that inaccurate data or poor algorithms within AI systems may go unnoticed during internal and regulatory reviews, potentially resulting in thousands of violations.

Amid this climate of skepticism, Sigal Mandelker, head of the Treasury Department’s Office of Terrorism and Financial Intelligence, said at the ABA/ABA Financial Crimes Enforcement Conference in Maryland on Monday that some banks have already used “innovative steps” to help identify front companies linked to Iran and North Korea.

“When responsibly deployed, these types of innovations are already proving invaluable,” she said.

Regulators indicated Monday they would seek input from industry on how new technologies have been applied, and also consider requests from financial institutions for limited forms of regulatory relief if doing so would enable them to experiment more easily.

But the benefits of such arrangements will depend on the details, Dan Newcomb, a New York-based attorney with Shearman & Sterling, told moneylaundering.com.

“It’s always good to know that your regulator is at least broadly in favor of innovation,” Newcomb said. “If some banks develop what they think of as a labor-saving, innovating AI tool, there will be a lot of skepticism I suspect from the regulators on the specifics.”