Agentic shopping raises liability gap no bank or retailer has solved

As AI shopping agents go mainstream, banks, retailers and card networks face unresolved questions over who bears the cost when software buys wrong.

By Disrupts Editorial Team Tue, 13 October 2026 AI level 2 of 5

AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean

A brightly lit data center aisle is lined with server racks, where colorful bundles of network cables arc across the floor and connect, with glowing data streams converging and branching out from the center.

AI shopping agents are moving from novelty to commercial reality just as the legal and financial infrastructure needed to support them lags badly behind. The question of who pays when an autonomous agent buys the wrong item, wrong size, wrong model, wrong price tier, is now surfacing at the highest levels of global finance. At Sibos in Miami earlier this month, Federal Reserve Governor Christopher Waller put the liability question directly to a room of senior bankers. Nobody had a clean answer.

That vacuum matters more than a single misplaced shoe order. Agentic commerce sits at the collision of three sectors: payments infrastructure, retail e-commerce, and AI platform strategy. The rules that govern each were written for human buyers, and none maps cleanly onto software acting on a shopper's behalf.

A payment approval is not proof of intent

The core problem is one of verification. A bank can authorise a payment without knowing whether the agent executed the shopper's actual instruction. Authorisation has always confirmed that funds exist and fraud screens were passed; it has never confirmed that the goods match the buyer's intent. When the buyer is software, that gap widens dramatically. Chargeback rules, the consumer protection backbone of card networks, were designed for disputes between a human cardholder and a merchant. Inserting an autonomous intermediary into that chain raises immediate questions about standing, evidence, and fault allocation.

Retailers are already rewriting the rules unilaterally. Target has updated its terms of service so that purchases made by agents it explicitly approves, within spending limits the shopper sets in advance, are treated as the shopper's own transaction. That approach transfers liability back to the consumer, sidestepping the question of what the agent actually did. Amazon and Perplexity, meanwhile, are in active litigation over agentic shopping; Amazon has separately moved to block Meta's shopping agent from its platform entirely, a signal that platform owners are treating agent access as a competitive and legal battleground rather than an interoperability question.

Phil Bruno, Chief Strategy and Growth Officer at payments software company ACI Worldwide and a former co-leader of McKinsey's Global Payments Practice, argues that agentic systems are only viable at scale with "permission that can be verified, identity that persists across the entire transaction, and evidence that ensures a fair outcome when something goes wrong." That three-part standard, verifiable permission, persistent identity, auditable evidence, does not yet exist as a common standard across card networks, platforms or retailers.

Consumer trust and the holiday crunch

The commercial stakes are immediate. ACI Worldwide and YouGov surveyed more than 1,000 US fashion shoppers in June 2026; 52% said they were not comfortable letting an AI assistant make purchases on their behalf at all. With the holiday shopping season weeks away, retailers and payment processors face a near-term collision between consumer wariness and platform-level ambitions to deploy shopping agents at scale.

For macro investors and cross-sector strategists, the liability gap is also a signal about the broader infrastructure deficit of the agentic economy. Visa and Mastercard are understood to be working through which agents gain payment credentials and on what terms, but no public framework has been published. That regulatory and standards vacuum is itself a strategic opportunity: whoever establishes the verifiable-identity and audit-trail layer for agentic commerce, whether a card network, a payments software incumbent, or a new entrant, captures a toll-road position across every sector that deploys AI purchasing at scale. Retail is the first visible test case. Procurement, travel booking, and subscription management are the next wave. The liability question Waller posed in Miami is not a payments edge case; it is the foundational governance question of the agentic economy.