Following Anthropic’s launch of new AI commerce agents alongside Visa and Mastercard, Monica Eaton, Founder and CEO of Chargebacks911, warns that the industry is racing to automate the purchase journey without resolving who carries the risk when an AI-driven transaction is later disputed:
“Everyone is focused on giving AI permission to buy but far fewer people are asking what happens when the customer later says: ‘That isn’t what I told it to buy.’
“That is where the real risk starts for merchants. They already face enormous difficulty distinguishing genuine fraud from first-party misuse and agentic commerce potentially adds another layer of plausible deniability. A consumer may have authorised an AI agent to act, but that does not necessarily mean they intended every individual product, price or merchant that the agent ultimately selected.
“If the industry gets this wrong, we could create an entirely new category of friendly fraud. Consumers will be able to argue that the agent misunderstood them, exceeded its instructions or made a purchase they never specifically approved, while the merchant is left trying to prove what happened after the fact.
“That cannot become the default model. If AI is going to make autonomous purchasing decisions, every transaction needs a clear and auditable record showing what the consumer instructed, what parameters were set, how the agent interpreted them and what was ultimately authorised.
“Merchants cannot become the insurer of every misunderstanding between a consumer and their AI.
“Agentic commerce has huge potential, but the dispute infrastructure has to evolve alongside the checkout experience. Making it easier for AI to complete a purchase without giving merchants the evidence needed to defend a legitimate transaction would simply shift more risk onto the businesses already carrying too much of it.”
