Artificial intelligence as a transaction monitoring tool … Getting to grips with the money laundromats … The algorithm is king … Machines are better than humans at risk identification … What could possibly go wrong? … Ariana Haghighi on the technological edge
It comes as no surprise, then, that the machine-learning tool’s next victim is the finance sector.
According to the United Nations Office on Drugs and Crime, the amount of money laundered each year globally is estimated at two trillion USD.
Consequently, the Australian Transactions Reports and Analysis Centre (AUSTRAC) has cracked down on anti-money laundering (AML) requirements for businesses.
Earlier this month, the Federal Court delivered its judgment in the Crown matter, where Crown Melbourne and Crown Perth are faced with fines totalling a $450 million, in part due to failures to monitor billions of dollars in transactions, including international payouts, in contravention of the Anti-Money Laundering and Counter-Terrorism Financing Act.
Over recent years, duties of financial institutions have expanded and penalties have doubled, all in an effort to improve transaction monitoring compliance.
Transaction monitoring is a colossal task, and illicit behaviour is becoming increasingly sophisticated and evasive. Every financial institution devotes significant time and budget to refining their program, and for large businesses with millions of transactions to parse, this job is a wild-goose chase.
Enter AI, which promises a rapid and systematic method for transaction monitoring.
Last month, Google announced the launch of an anti-money laundering AI-powered product for financial institutions. This product, which follows a customer risk-based programme rather than rules-based, aligns with recent AUSTRAC guidance on anti-money laundering programme methodology.
Google’s machine purports to generate a customer risk score by synthesising data on transactional patterns, network behaviour and Know Your Customer (KYC).
Many financial institutions have followed Google’s lead, including HSBC, an unwitting assistant of the 2017 Global Laundromat scandal.
That operation connected UK banks to an illicit Russian money laundering operation which created fictitious companies exchanging loans.
Since HSBC’s complicity in the impropriety leading to $US740 million in losses, the bank has faced pressure to tighten its anti-money laundering procedures.
These intelligent AML systems tout improved precision with risk detection and high efficiency, but to what extent will the use of AI change anything?
In 2021, the Financial Action Task Force (FATF) published a paper exploring the opportunities, as well as challenges, presented by AI-led AML.
This paper argues that AI’s greatest strength is risk identification, which is often at fault in human-based systems. With AI, risk analysis can be more dynamic, as a customer risk profile can be updated automatically based on changing data rather than re-assessed during periodic audits.
The FATF also envisions a more financially-inclusive future, where digital identification poses benefits for both entities and individuals.
For financial institutions, digital systems support cheaper and faster onboarding processes, as they can allow for wider data sets, encompassing customers without a traditional credit record.
The portability intrinsic to digital systems also offers customers faster ways to engage with verification, such as through their smartphone.
The capacities of AI extend beyond typical digital technologies, as the algorithm tweaks itself each time it repeats a task. Whilst self-improving, some experts fear this could lead to inconsistencies, or a world where we are beholden to the algorithm.
The FATF report contemplates AI playing a beneficial role in transaction monitoring. Time will tell whether AI resolves also poses problems of its own.
After all, the automated online compliance initiative (i.e. Robodebt) is still fresh in our minds.