The regulators are curious again.

Last week, the Office of the Comptroller of the Currency (OCC), the Federal Reserve Board and three more regulatory bodies, put out a call for public input on how artificial intelligence (AI) and machine learning (ML) are being used in the financial services sector, and the risks they pose.

The stated purpose for seeking public input was to determine whether new rules or guidance “would be helpful for financial institutions’ use of AI in a safe and sound manner and in compliance with applicable laws and regulations,” an OCC spokesperson told Bank Automation News.

The 17 questions raised by the Federal Reserve and OCC together with the Federal Deposit Insurance Corporation (FDIC), Consumer Financial Protection Bureau and the National Credit Union Administration (NCUA) also offer insight into how the watchdogs are looking at emerging technology.

To learn what banks, tech companies, and bankers’ associations think about use cases for AI in financial services, BAN reviewed public comments received by the OCC in response to an advance notice of proposed rulemaking (ANPR) the agency put when it was “reviewing its regulations on bank digital activities,” in July 2020.

Although AI and ML can be leveraged to improve compliance, security and credit underwriting, the regulatory ambiguity on risk models and explainability issues surrounding AI and ML could be slowing down adoption, according to comments shared by groups like the American Bankers Association (ABA), Bank Policy Institute, and banks like PNC and U.S. Bank.

Explainability in AI refers to the extent to which internal mechanics of a system can be explained in a way that allows those without a tech background to understand how the AI system arrived at a decision.

Opportunities, risks and clarifications needed

“AI can help banks extend credit to more borrowers, enhance the customer experience, improve fraud detection, lower the cost of offering services,” noted the comment filed by the ABA, a banking trade association that represents banks of all sizes and roughly 2 million of their employees.

While existing banking regulations cover much of what AI is being used for, the ABA urged the OCC to consider clarifying certain rules to facilitate the use of such technologies, particularly with regard to model risk management, fair lending, cybersecurity, and data privacy. “The use of AI does not present novel risks but may amplify certain existing risks that must be managed,” the ABA commented.

The $533 billion U.S. Bank uses AI and ML techniques for marketing, fraud detection and anti-money laundering screening, according to bank’s comment in response to the ANPR. Although the bank is exploring “the feasibility of these methodologies for credit underwriting,” concerns about transparency and explainability have slowed down adoption in the space, the comment noted.

The OCC should communicate with the industry on the permissibility of these tools and on the “standards that will be used by examiners in evaluating these methods,” according to U.S. Bank. It added that the regulator should also “clarify fair-lending expectations with respect to AI-based models.”

In addition to guidance on how banks use new tools, some of the comments also noted how a data-led approach in banking also creates opportunities for financial institutions to become data processors and even white-label sellers of financial software.

“The OCC should expand and clarify the national banks’ authority to engage in data processing and data transmission services,” said the comment filed by the Bank Policy Institute, a financial services’ lobbying and advocacy group that represents the leading U.S. banks. It added that there are constant challenges to ensuring transparency and controls around how vendors build their AI products, due to “vendors’ reluctance to share” algorithms because of intellectual property and competitive concerns.

The Institute also commented that the OCC should work with the Consumer Financial Protection Bureau to ensure that both banks and non-banks apply consistent and robust risk management and controls to their AI tools in order to mitigate risks like data quality and privacy issues.

Like most of other banks and vendors that responded, the $433 billion PNC Bank in its filed comment urged the OCC to maintain its “technology-agnostic” approach towards regulation, meaning that all banking services should receive the same regulatory treatment, regardless of whether they’re performed on paper or electronically.

“We encourage the OCC to ensure that banks have the flexibility to move forward with pilot programs that present limited risk and are structured to ensure compliance,” PNC Bank’s comment noted, adding that the programs would allow the banks and regulators to better understand whether use of the new offering is consistent with an institution’s risk appetite and business strategy.

As regulators await fresh inputs on how AI and ML are being used in the financial services sector, the comments already filed with the OCC indicate that the regulation-heavy nature of banking means the adoption of AI and ML could be significantly impacted.

For now, the questions raised by five regulatory bodies point to the fact that the watchdogs are keen to make sense of the role of emerging automation techniques in the financial sector.