AI agents expose banks to new risks

Por Victoria Guida, Jasper Goodman and Aiden Reiter — POLITICO – TOP Stories

The age of the AI personal assistant is here, courtesy of Meta, and it could bring about a reckoning for the global financial system.

It’s the latest wake-up call for banks and their regulators about the dangers posed by a swiftly emerging wave of transformative technology.

The launch this month of Meta’s Muse has brought new attention to the possibility that people might use so-called AI agents to easily shift their money into competing lenders that pay more interest on deposits, draining a cheap source of funding for many banks. While that could be a boon for everyday people, it’s a potential stability risk for firms that sit at the center of the U.S. economy.

That risk threatens to stoke tensions between two of Washington’s most powerful lobbying forces — AI firms and big banks — and expand the scope of the battlefield for Meta, which is already facing pushback from Amazon. But in the meantime, financial institutions, regulators and lawmakers are still in the early stages of grappling with how the technology might affect depositor behavior, as well as how to address a slew of other questions posed by AI agents, including who bears legal responsibility for their behavior and what consumer protections might be needed.

“There’s a whole class of worries there that I’ve been worried about for a while,” Rep. Bill Foster, a senior Illinois Democrat on the House Financial Services Committee, said in an interview.

Rapid developments in AI have led to a flurry of activity in Washington as policymakers consider a spectrum of possible implications from the technology, from catastrophic cybersecurity dangers to potential productivity gains. Today at the White House, President Donald Trump and House Speaker Mike Johnson will meet with Meta CEO Mark Zuckerberg and other top executives on matters relating to AI.

Foster, who is in line to chair a House Financial Services subcommittee on financial institutions if his party retakes the House, said he met last week with bank regulators who are interested in areas where Congress might look to legislate.

“Dealing with all the challenges with agentic AI was probably the top of my list,” he said.

Meta, the Federal Reserve and the FDIC all declined to comment. The Office of the Comptroller of the Currency did not respond to a request for comment.

Jonathan McKernan, who recently left his job as Treasury’s domestic finance chief, said on IntraFi’s “Banking With Interest” podcast earlier this month that banks have benefited from how difficult it is for customers to switch where they put their deposits. Banks often offer depositors less than 1% interest on savings, while some nonbank firms can provide rates upwards of 4%.

“Banks are a lot more than that, especially the community bank; you have the relationship with your borrowers, you have trust, you provide a lot of services,” he said. “But the frictions [associated with moving deposits] are part of the business model, and these innovations that we have now are smoothing away those frictions.”

Ethan Bloch, who handles personal finance for OpenAI, pushed back on the idea that it would radically shift the calculus for banks.

“Consumers want a name brand they can trust with their life savings,” he said in a post on X. “These are the 2 biggest reasons Chase has $1T in deposits even though 98% of it could be earning 350x the yield. Agents don’t change any of this.”

But former Democratic Rep. John Delaney, the founder and executive chair of Maryland-based Forbright Bank, said the shift will force banks to step up to avoid agents directing deposits out of their accounts. “The banking system has to effectively reorganize their business model to be profitable while paying people a fair rate, not by underpaying them,” he said.

Banks are wary of the challenges the technology could pose to their business models but are also hesitant to make any firm predictions: AI could threaten their bottom lines, but it could also boost profits through new offerings and novel ways to attract and retain depositors. And stepping ahead of regulators makes some cautious.

“Banks of all sizes are always adapting and adjusting to risks to their deposit base,” said Sarah Grano, a spokesperson for the American Bankers Association. “The specific risks posed by agentic AI, including faster and automated money movement, highlight the need for prudent regulatory guardrails to support responsible financial innovation.”

Consumer Bankers Association President Lindsey Johnson in a statement cited questions about how anti-money laundering rules apply to AI agents.

“Consumers should be able to benefit from these technologies without weakening the safeguards that protect them and the financial system,” said Johnson, whose group represents retail banks.

Sen. Mike Rounds (R-S.D.) said regulators need to make sure they have the tools to properly oversee AI agents “so that they can respond more quickly.”

“The only way that you’re going to protect against agentics from moving very rapidly is to have the oversight tool that can move just as rapidly,” he said.

For their part, regulators are considering the implications posed by bank customers using AI agents but are closer to moving forward on policies related to how banks themselves use AI agents. Specifically, they are considering issuing a request for information on how the industry is using the technology and where rules might make sense, according to an agency official.

Former acting Comptroller of the Currency Michael Hsu, who is now co-chair of the financial services workstream at nonprofit engineering consortium MLCommons, said regulators should have two immediate priorities in providing more guidance to banks. One is making sure that AI agents used by banks are “bounded,” or have parameters that they provably will not work outside of. The second is incident reporting.

“It’s a little bit like cyber incident reporting, like when things don’t go as planned,” Hsu said. “There’s no [National Transportation Safety Board] for AI. There absolutely should be one, and it’s got to be a little bit different than NTSB because these incidents happen way more frequently, and they’re just weirder.”

Sam Sutton contributed to this report.

Fonte: POLITICO – TOP Stories