Big Banks Say They're Uneasy About People Shopping via AI Agents

Six major global banks, including Bank of America and Capital One, warned that autonomous shopping assistants - AI agents that can search, choose and pay for purchases on behalf of users - pose serious consumer and industry risks as they gain autonomy. Consumers remain reluctant to hand over broad access to funds and accounts, fearing mistakes, overspending or scams, and some industry figures doubt these agents will fundamentally change shopping beyond improving information. Banks say adoption is growing among AI developers, retailers and payment processors, but trust is low because agents can act opaquely and in ways users don’t expect.

The banks’ report, titled “Building Trust in Agentic Commerce,” outlines five categories of concern - transparency, safety, privacy and data, choice, and interoperability - and warns of specific threats: agents could steer users toward higher-commission products or cheaper-to-run options, increase fraud and dispute rates, and magnify damage from data breaches. Recent incidents underline those dangers, such as a zero-day flaw in Meta’s Muse assistant and Amazon’s complaint that the agent failed to identify itself while shopping. The banks don’t dismiss agentic commerce’s potential to go mainstream, but insist industry standards, consumer protections and technical safeguards must be established; a follow-up paper will propose ways to implement the five key principles.