Reference no: EM133047369

HSBC Adopts Machine Learning Artificial Intelligence to Fight Money Laundering

HSBC Holdings is a British multinational investment bank and financial services holding company. With a history dating back to 1865, HSBC first opened its doors in Hong Kong where it hoped to finance trade between Europe and Asia. Its name comes from a company, formed in London by the Hong Kong and Shanghai Banking Corporation (HSBC), to act as a new group holding company in 1991. For more than 150 years, HSBC has prospered through many kinds of changes-revolutions, economic crises, and new technologies. Today, HSBC is Europe's largest bank, and the seventh largest bank in the world. It serves more than 40 million customers through its global business and has $2.251 (€2.055) billion in assets.

Money Laundering at HSBC

Despite HSBC's prominence in the banking world, its record has not gone untarnished during the past two decades. On four different occasions (in 2003, 2010, 2012, and 2015), the banking organization was cited for lax anti-money laundering practices by regulators in the United States, India, Argentina, Switzerland, and the United Kingdom. Not only did the bank fail to recognize the money laundering activities using its banking accounts, but in several situations, officers and employees of the bank played a direct role in money laundering, tax evasion, and other financial crimes. These lapses resulted in more than $1.94 billion in fines, a penalty that many regard as shamefully low, considering that HSBC's money laundering was connected to nuclear weapon development programs in Iran and North Korea, financing of terrorist organizations, and large-scale drug trafficking operations linked to fraud, organized crime, and murder.To rehabilitate the bank's reputation for lax regulation and to discourage use of its operations by criminal and terrorist organizations, current HSBC executives have embarked on new efforts to detect and eliminate money laundering and tax evasion schemes. To help with their renewed commitment to fight money laundering activities and fraud, the bank has contracted with companies like Ayasdi, a Silicon Valley-based developer of machine learning artificial intelligence (AI) programs. Because of HSBC's size and extensive global network of banking operations, employees, and customers, traditional methods of monitoring transactions and account fluctuations for indications of illegal activity are inefficient given the sheer amount of financial data generated daily.

Anti-Money Laundering (AML) Applications

Traditionally, banks employed myriad rules that act as filters for spotting suspected activity. For instance, consider an account with an average monthly balance of $3000, and an average transaction amount below $250. If that account suddenly has a deposit of $20,000 followed by a transfer of these funds to an international account, the situation should raise flags. However, when transactions like this are buried in a seemingly endless sea of financial activity and dependent on human auditors to find these cases, it is likely that a great deal of questionable activity will go unnoticed, even when bank officers and employees are doing their best to identify suspicious activity. Anti-money laundering (AML) applications powered by AI have a much better chance of catching situations like this as well as more nuanced attempts to evade money laundering regulations.

AML Application Powered by AIAs a first step in the development of an AML solution, an Ayasdi data scientist used feature engineering to identify elements in HSBC's data that would be helpful training an AI application to identify cases of suspected money laundering activity. Then machine learning programs developed by Ayasdi applied several algorithms that sifted through HSBC's data to identify numerous patterns or clusters of financial activity. Ayasdi staff then worked closely with the bank's IT experts and a team of modeling specialists from HSBC to review and validate what the AI algorithms were predicting so that the patterns and clusters uncovered by the AML program could be explained. Because of the tight regulatory nature of the banking industry, it isn't enough to just have the AI point out data patterns or clusters related to fraud (an outcome sometimes called the "black box" effect of machine learning). Instead, bankers must be able to explain how the patterns were derived and why they are indicative fraudulent activity.

Results of Ayasdi AML Solution

In the end, the Ayasdi AML application was able to uncover several new patterns in the data directly related to new fraud cases as well as reduce HSBC's incidents of false positives (cases that would have been flagged as fraud, but were not really linked to illegal activity) by 20%.

Future of AI Applications in the Banking IndustryIn addition to the use of machine learning AI to combat money laundering behavior, financial organizations today are using machine learning applications to uncover other types of criminal behavior such as fraudulent credit card transactions, and detection of unusual activity on a bank's computer system that might indicate penetration by a hacker or someone using an employee's account without permission. HSBC officers in charge of financial crime mitigation are actively working to harness technology and data with the goal of identifying criminal activity in real time. It is expected that the banking industry will embrace AI banking programs for many of the same reasons that banks have been rapid adopters of mobile technology in the consumer banking sector. AI promises to help banks enhance customer convenience, trust, and security, all of which are critical strategies for differentiating themselves from the competition.

Questions

1. Why would a bank like HSBC find it difficult to identify transactions and account fluctuations that should be clear signs of money laundering or other fraudulent behavior?

2. How did Ayasdi's machine learning AI help HSBC uncover cases of money laundering?

3. What is the purpose of "feature engineering" conducted prior to submitting data to the machine learning application?

4. What is the black box effect of machine learning applications?

5. Why are banks likely to adopt different kinds of AI technologies in the future? What do they hope to gain from this emerging area of technology?

Sources: Compiled from Irrera (2017), Arnold (2018), Faggella (2018a), Symphony Ayasdi (2019), and Worldwide Business Research (2020).