AMLC to banks: Show AI helps curb financial crime

MANILA, Philippines — Banks and other financial institutions using artificial intelligence to fight financial crime should be able to show that the technology helps detect risks and that its decisions could be examined, according to the Anti-Money Laundering Council.

AMLC executive director Ronel Buenaventura said the agency’s supervision would increasingly draw on data and intelligence as criminals make greater use of digital channels. For institutions adopting AI, the test is whether it produces useful results beyond speeding up routine compliance work.

“You can use AI to automate processes, you can use AI to generate insights,” Buenaventura said. What matters, he added, is whether those tools help institutions prevent or detect financial crime.

Buenaventura said institutions should be able to explain and audit how their AI systems work, from the data used and the way models are checked to the decisions they produce.

Human oversight remains necessary, while boards must receive and act on significant risks identified by the technology.

The AMLC has likewise begun using new tools in its own supervision, he said. Its assessments and audits will become more data driven as financial institutions expand their use of technology.

Maya chief risk and compliance officer Julie Reyes said financial institutions need to examine fraud and money laundering together because proceeds from scams can move quickly through digital accounts and payment channels.

“What we have done is come up with an integrated approach, like several years back,” Reyes said. She added that fraud and anti-money laundering teams need to understand each other’s work and that their systems must be connected.

Reyes said Maya has formed a team to validate the models it uses and has brought in data scientists who can assess risks in the context of the business. The company is also working on an inventory of its AI models, she said.

As those models multiply, institutions need to monitor whether their accuracy changes over time and report relevant measures to their technology governance committees, Reyes said. Decisions with a significant effect on customers require particular care.

Reyes said institutions must be able to explain decisions made with the help of AI, particularly when they affect customers. “You decide to decline a client, why? So, those things are very sensitive and it has to be explainable.”

Buenaventura said the aim is to make financial crime controls more effective as risks evolve.

“At the end of the day, it’s actually how effective the institution is in combating or preventing or mitigating the risk,” he said.