Fed's Bowman: AI Rules Must Not Block Financial Inclusion

Federal Reserve Vice Chair for Supervision Michelle W. Bowman wants the central bank's supervisory approach to artificial intelligence to leave room for smaller lenders to innovate. In a pre-recorded speech Tuesday to the Fed's Financial Inclusion Conference, she said AI can expand credit access to the unbanked and underbanked by giving banks better tools to assess creditworthiness.

Bowman acknowledged the tension. AI that directly affects individual lending decisions raises legal compliance questions. The goal, she said, is to support responsible innovation without stifling it.

"That starts with greater clarity on what level of oversight is appropriate for different AI applications," Bowman said.

She called for supervisory guidance that does not block smaller banks from adopting modern technology. Lenders should be able to implement AI in ways that fit their specific business models, she said, and lower-risk uses should face lighter oversight.

"Financial institutions should leverage their existing risk-management frameworks, adding appropriate enhancements and controls tailored to the specific risks that each AI application presents," Bowman said.

Bowman chairs the Financial Stability Board's Standing Committee on Supervisory and Regulatory Cooperation. The FSB published a consultation report in June that outlines sound practices for AI governance, risk management and third-party oversight. The comment period closes July 22.

"That report also reflects our broader commitment of maintaining an ongoing dialogue between bankers and supervisors to ensure our approach keeps pace with innovation while safeguarding safety and soundness," Bowman said. "We have been engaging with banks on AI for nearly a decade, and as use cases expand and technology evolves, that conversation becomes even more important."

The speech lands as financial services firms embed AI into revenue recognition, credit scoring and sales forecasting. A PYMNTS Intelligence report found the sector has chosen to deploy AI "when outcomes are certain and the consequences of error are manageable."

For smaller banks, the practical question is whether the Fed's examiners will treat a credit-scoring model built on a large language model the same way they treat one built on logistic regression. Bowman's answer Tuesday suggested a tiered approach: heavier scrutiny for AI that directly determines whether a borrower gets a loan, lighter touch for AI that flags suspicious transactions or automates back-office tasks.

The distinction matters because the unbanked and underbanked populations are precisely the ones most likely to lack the traditional credit history that conventional scoring models require. AI models that incorporate cash-flow data, utility payments or rent history could open credit to those borrowers. If the Fed's supervisory stance forces banks to treat every AI application as equally risky, smaller lenders may simply skip the technology altogether.

Bowman's comments suggest she wants to avoid that outcome. The FSB consultation report, and the comment period now open, give the industry a chance to shape how the rules take shape.