Despite enormous investments in Anti-Money Laundering (AML) detection - from advanced monitoring systems to large compliance teams - money laundering remains alarmingly persistent. Criminals stay several steps ahead, while financial institutions find themselves trapped in a game of "compliance theatre".
AML detection is inherently difficult: money launderers are agile, adaptive, and operate across borders and institutions. But the real challenge lies not just in complexity - it’s in the misaligned incentives and checkbox mentality that dominate the system.
Financial institutions are required to:
Implement a predefined set of detection rules (which are often public knowledge among criminals)
File Suspicious Activity Reports (SARs) when certain patterns or thresholds are triggered
The cost of AML is staggering compared to its measurable benefits. For example, in the Netherlands, it’s estimated that around 20% of bank employees - roughly 13,000 people - are engaged in AML activities, costing banks approximately €1.4 billion annually. Meanwhile, only about €400 million in criminal funds was seized in the same year. A poor return from a societal perspective. And these costs keep increasing due to tighter regulations and higher fines.
Yet calling for fewer AML controls isn’t the solution. The scale of illicit funds flowing through the financial system is massive - estimated at $2 trillion annually - fueling drug trafficking, weapons trading, human exploitation and other severe criminal activities. AML efforts must be strengthened, not abandoned. Banks are the first line of defence.
But current approaches are failing to catch terrorists, criminals, and insiders. Automated transaction monitoring has ballooned into massive operations involving costly staff and consultants.
Many institutions are now pinning their hopes on AI to improve detection and cut costs - particularly by reducing compliance headcount. But AI is only as effective as the data it’s fed. It needs high-quality feedback, including clear distinctions between true criminal cases and false positives.
As long as banks follow procedures and file SARs, they’ve technically met their compliance obligations - even if those reports don’t result in convictions.
The AML system can only become effective when core weaknesses are addressed:
No penalty for over-reporting: Banks face fines for under-reporting but not for over-reporting. This encourages them to flood authorities with thousands of SARs - often low-quality and repetitive. Law enforcement is left searching for needles in an ever-growing haystack. In the UK, over 872,000 SARs were filed in 2023–2024 (up 52% since 2020), while the US saw 3.8 million filings in the same period.
No feedback loop: SARs are processed confidentially. Banks aren’t told whether their reports were helpful, if the account was part of a criminal network, or whether the customer relationship should be ended. Without this feedback, banks cannot improve their detection models.
Misaligned incentives: Banks are focused on avoiding fines and reputational damage - not on stopping crime. The result is a system that values paperwork over prevention.
If the goal is to comply, then the system is working. But if the goal is to stop crime, it’s not fit for purpose. Filing SARs is incentivized. Finding money launderers is not.
To truly fix AML, we must realign incentives so that stopping financial crime becomes a shared goal. Here are some ideas:
Reward financial institutions for successful detection: When a SAR leads to a conviction or asset seizure, a portion of the recovered funds could be returned to the reporting bank. Conversely, clearly incorrect (over-reporting) or negligent SARs could result in fines - encouraging quality over quantity.
Create a closed AML fund: Fines and recovered criminal assets could be pooled into a reward fund that compensates banks for effective SARs. The system becomes self-sustaining: more detection leads to more rewards.
Traceback analysis after conviction: For each crime conviction, regulators could trace the money trail. Banks that raised flags are rewarded; those that missed obvious signs could be fined. This fosters accountability and continuous learning.
Controlled feedback loops: Even if detailed case data must remain confidential, regulators could share anonymized insights: Which SARs were most helpful? Which indicators proved predictive? This would help banks improve model accuracy and reduce false positives.
Encourage cross-bank collaboration: Criminals operate across banks, countries, and systems. AML cannot succeed if each institution acts alone. Privacy-preserving data-sharing (e.g. federated learning or homomorphic encryption) could reveal hidden patterns invisible to individual banks.
Leverage AI and behavioral intelligence: Rule-based systems are easily reverse-engineered. AI-driven, pattern-based models - trained with quality feedback - offer adaptability and precision.
Today’s AML framework rewards compliance, not outcomes. We measure the number of SARs filed, not the criminal money stopped. Regulators must shift to outcome-based supervision - where success means actual impact, not just paperwork.
Banks and regulators want the same thing - a safer financial system - but operate under opposing incentives. Realigning those incentives is the only way AML can truly work.
Comments
Post a Comment