What Muse said and why the legal line matters
A prompt to Muse, Meta Platforms' personal agent, asked whether it acts as a user's legal representative. It responded, "Functionally yes, legally not quite," and explained it does not owe a fiduciary duty of care "the way a lawyer or financial adviser would." The question came from the Bank Policy Institute, a banking lobbying group, and Wall Street now wants precision on what that "not quite" means in practice and who covers any losses if an agent goes off script.
Meta released Muse in September, which sped up industry efforts to support agent-driven tasks. More than 10 million people in the US and Canada tried Muse in its first six weeks, Mark Zuckerberg told analysts. Looking ahead, Bloomberg Intelligence estimates agent-executed volume could reach 470 billion dollars by 2030, according to a September note by Diksha Gera and Sena Ustuner.
Why banks and payment firms are hitting pause
Consumers are increasingly letting AI agents access their bank and card credentials. That raises the specter of bots that can shuttle money, shop, open new credit, or refinance - by design, with user permission. The worry is not theoretical. At JPMorgan, the number of consumer payments initiated by agents has more than quadrupled since Muse launched, though the dollars moved remain small.
Liability is the sticking point. "Before this technology scales to hundreds of millions of Americans, we need to get some fundamental protections right," said Cam Fowler, chief executive of Zelle owner Early Warning Services. To curb exposure, several firms have followed suit: JPMorgan CEO Jamie Dimon said the bank is blocking these agents on the consumer side, while Charles Schwab and American Express have done the same, with Royal Bank of Canada joining later. Retailers, including Amazon, have also started shutting the door.
What early usage looks like and who is at risk
Not every bank is slamming the brakes. M&T Bank's finance chief, Daryl Bible, said on Nov. 5 the bank has seen bots pull basic data, like balances, but not shift funds or request product or service changes. "We're closely watching it right now and if we see anything that we think is not appropriate, we will try to shut that down," he said, adding they are not currently seeing problematic behavior.
AI agents can already surface money tips, from higher rate savings options to cheaper insurance. But blunt instructions can backfire. An agent that chases a lower premium might cut coverage too far, or one that parks cash in a better yielding account without seeing a rent bill coming could trigger an overdraft. Brian Shearer, a consumer protection attorney, warned that if attackers use AI agents, and if AI firms collect bank-derived data, those companies could end up holding "the most sensitive and important information about a large group of people." Shearer, who heads the Vanderbilt Policy Accelerator's competition and regulatory policy work, said he worries about firms stumbling into mass data collection.
When developments raise complex questions about responsibility, stay informed with our free Market Briefs newsletter that explains it simply
The policy hole and the path to scale
Banks and payments companies had been working through the plumbing needed for "agentic" transactions, where AI carries out tasks for a customer. Then Meta shipped Muse, and the timeline sped up faster than many expected. At the same time, the Consumer Financial Protection Bureau finalized a rule during the Biden administration to govern how consumers grant outside firms access to their financial data, prompting swift lawsuits from bank trade groups. The Trump administration later eliminated that rule, leaving a wide-open space just as agent-driven activity took off.
Former CFPB Director Rohit Chopra said banks passed on a chance to join a federal framework for consumer-permissioned data and instead fought it, which left today's landscape where consumers hand passwords to AI companies to move money "in a more methodical way."
Bloomberg Intelligence sees agentic AI reshaping how customers pick banks well before it hits deposit pricing, with Oliver Wyman's Aaron Fine calling customer acquisition the nearer term risk. Adoption skews young, with only 4 percent of consumers 55 and older showing interest in autonomous-agent transactions versus 40 percent among younger users. And while hunting for better yields is not new, agents could make shifting surplus deposits far easier. That is why many banks are limiting agent access over security, authorization, and liability concerns.
The bottom line for investors
User demand is building, guardrails are not, and the liability question sits in the middle. Banks that restrict agents may slow outflows and fraud risk, while institutions that enable them could win younger, yield-seeking customers. As usage climbs - remember, JPMorgan's agent-initiated payment count surged after Muse - watch who sets the rules, who blocks access, and who builds safe pipes. That is where the deposit mix, fee exposure, and customer growth will move first.
Because clarity matters as technology and finance intersect, sign up for Market Briefs to get the essentials in minutes a day
