US regulators take interest in AI in banking

A group of US banking regulators—including the Federal Reserve, the Federal Deposit Insurance Corporation (FDIC), and the Consumer Financial Protection Bureau (CFPB), among others—has issued a statement that they’re seeking public input on the rising usage of AI by financial institutions (FIs), Reuters reports. They specified that they want feedback on “how financial institutions use AI in their activities, including fraud prevention, personalization of customer services, credit underwriting, and other operations.” The request for information also seeks to understand whether any clarification from the regulatory agencies would help FIs use AI in a safe way that complies with laws and regulations.

Regulatory scrutiny of AI in banking could add to the perceived risk that banks take on when they deploy the technology in their businesses. The use of AI in banking inherently carries risks, chief among them being:

  • AI models run the risk of becoming 'black boxes.' Because machine learning algorithms are so complex, their decisions can be tough to analyze or explain, even by the staff whose job it is to design and maintain them, making accountability and error detection a challenge.
  • Flawed training data can create unintended biases in AI models. AI algorithms that are trained on incomplete, biased, or extraneous data can yield judgments that are also biased, causing a range of issues from inconvenient technical errors to inadvertent discrimination.
  • The vast amount of data AI applications require can make managing and communicating the use of that data difficult. AI models draw on large pools of data—and banks have to use it in ways customers don't find intrusive, as well as explain why they need the information in order to avoid damaging customer trust.

And any trepidation that banks may feel regarding the deployment of AI may only be heightened by the sense that government bodies have started paying close attention to the presence of AI in the banking industry.

But the exploratory interest of regulatory agencies is also a major opportunity for banks and regulators to align on the benefits, pitfalls, and expectations for AI in the sector. If FIs can clearly communicate the ways they use AI, the benefits it offers the banking industry, and what happens when something goes wrong with the technology, then those agencies can regulate AI in financial services in a more informed way. More understanding regulators will be significantly better for banks in the long run, especially when an issue with an AI model occurs that leads to a problematic result, such as an unintended bias or data leakage.