sdecoret - stock.adobe.com
Using AI in AML fights fraud while protecting privacy
Money laundering and fraud remain a risk for financial institutions, but AI can act as a useful tool against a constantly evolving financial enemy.
Despite widespread advancements in AI, privacy and protection, fraud and financial crimes always seem to persist. The advent of the internet and the ushering in of the digital age simply moved financial criminal activity online. It is estimated by the UN that the amount of money laundered globally equates to between 2-5% of the global GDP, or billions to trillions of U.S. dollars each year.
With regulators and countries fighting against money laundering, AI is being explored as a potential answer for banks and other financial institutions. Thousands of transactions can be sorted through with a trained bot at a significantly faster rate than with a human employee. Detecting fraud and money laundering activities has become another proving ground for AI, and with billions on the line, banks are betting on the technology.
AI makes fraud proactive
When it comes to risk and compliance, especially concerning fraudulent activity, companies need to be on watch. Fraudulent transactions and money laundering have been in business consciousness for decades. It was the rise of organized crime, tax evasion and money laundering that cemented those terms in the public mind -- and they continue to be a business detriment.
The use of AI in AML and fraud prevention is becoming increasingly relevant as other AI products and frameworks see business success, said Madhu Bhattacharyya, managing director and global leader of Protiviti's Enterprise Data and Analytics practice.
"There is a significant risk for companies when it comes to money laundering," Bhattacharyya said. "Any company -- whether it be retailers, banks, insurance companies, MSBs, casinos and so on and so forth -- where there is a transaction in the form of money, gift cards, real estate, etc., is exposed to the risk of financial crime."
Anti-money laundering policies and technology have been developed rapidly to combat the increased risk that comes from the digital world. Almost countless transactions occur online and finding fraudulent or illegally gained money is akin to the proverbial needle in the haystack. The scale of fraudulent activity continues to cause significant headaches to banks -- especially since they are scrambling to solve data breaches and laundered money.
"The world of risk and compliance, including anti-money laundering (AML) and fraud, has realized the importance of being proactive rather than reactive," Bhattacharyya noted.
With so much data to process and sort through, finding unusual activity can be difficult on its own. But discovering legitimate mischievous activity can prove a significant challenge without creating false positives during proactive financial transaction searches.
Introducing AI to AML
By utilizing machine learning, companies can compile data from transactions and train the machine to differentiate between a normal transaction and suspect activity. As the differences become clearer (as the machine is trained on more and more data) the machine can be set to work on thousands of transactions and bring suspicious activity to the attention of the bank.
"Using data to identify probable fraudulent activity even before it has happened reduces the risk of fraud and money laundering," Bhattacharyya said.
Unusually large transfers, known suspect accounts or out of the ordinary deposits can all be flagged automatically, and in mass quantity, by AI systems and sent to be reviewed. Online banking has brought convenience to consumers but also serious challenges when it comes to tracking criminal activity. AI has become another tool in the rule-of-law toolbox for banks and financial institutions to monitor suspicious activity.
Money laundering techniques are constantly evolving in order to bypass advancements in AML, making the job of regulators and institutions exceedingly more difficult. Turning to AI holds the potential to help them both as well as aid in building consumer trust. On top of this, the fines associated with noncompliance with AML regulations and fraud detection can potentially be avoided with the assistance of AI.
"Businesses, across geographies, want to gain a competitive advantage with the use of AI and its applications," Bhattacharyya said. Financial institutions are only one of many hoping to create niche AI strategies for costly problems.
Ensuring ethical AI
By training machines to help identify probable fraud, enterprises hope to build trust and limit damage.
"AI algorithms can be used to identify the uniqueness of a transaction, reduce false positives and optimize business scenarios," Bhattacharyya said. "Models are built, implemented, validated and monitored for deviations and relevance."
But lack of clarity with how the data is handled, the algorithms being utilized and the details of the AI can create privacy concerns. Financial institutions have been careful when it comes to adopting new technologies like AI when the privacy concerns have not been reasonably resolved.
The complexity behind the models can leave banks unable to explain how they comply with regulations. With intense scrutiny over privacy and compliance, banks have been hesitant to invest in the technology.
Proper supervision and understanding are key to the utilization of AI in fraud detection and AML. Through either supervised or unsupervised learning, it is important that humans are kept in the process. False positives are a danger that AI, as well as human counterparts, can create and having checks within the approach to fraud detection is necessary.
AI has the potential to further assist with AML and fraud detection when utilized in concert with existing programs and under human supervision.
