Six of the world's biggest banks just told the AI industry: your shopping agents are going to get scammed.
On Tuesday, Bank of America, Capital One, ASB Bank, Commonwealth Bank of Australia, ING Group, and NatWest Group published a paper called "Building Trust in Agentic Commerce." It lays out five principles - transparency, safety, privacy and data, choice, and interoperability - that the banks want AI companies to follow as agents start buying things on customers' behalf. The banks say shoppers aren't sure whether an AI agent will spend the right amount, buy the right item, or protect them if a purchase goes wrong. They also flag risky habits among some AI providers, like keying a customer's card details directly into merchant sites or defaulting to payment methods with weaker fraud protections.
The fine print here is really an insurance argument: banks eat the cost when agentic purchases go sideways, through chargebacks and disputes, and they don't yet trust that AI companies will build in the safeguards to prevent that. It's a familiar problem in a new shape - card networks spent decades writing fraud-liability rules for humans clicking "buy," and now that framework has to be rebuilt for software making the call instead.
The timing is pointed: the paper landed within a day of a reported zero-day bug in Meta's Muse assistant, and hours after Amazon said it would block Muse from making purchases on its platform altogether.