The plumbing for AI agents to buy things on your behalf is officially live. The people those agents are supposedly shopping for are getting more nervous, not less.
Visa rolled out its Trusted Agent Protocol earlier this month, and Mastercard launched a parallel framework of its own. Both are designed to let ChatGPT, Gemini, and Claude run real transactions through the card networks without a human ever clicking confirm. Visa's Chief Product and Strategy Officer Jack Forestell called the agentic web "the biggest opportunity that I've seen in my 20-plus years in payment technology."
Consumers do not appear to share the enthusiasm. A Q1 2026 Riskified survey found that 55% of shoppers are uncomfortable with AI agents making purchases on their behalf. The same survey one quarter earlier had 70% saying they were at least somewhat comfortable with the idea. As the rails actually came online, the comfort level moved the wrong direction.
The bigger issue is what happens when an agent screws up. When asked who should pay for an unauthorized AI purchase, 50.8% of respondents in the Riskified survey pointed at the AI platform itself. Only 23.2% blamed the retailer, and 18.7% were willing to accept personal responsibility. Roughly half the country already thinks the AI platform owes them a refund if an agent buys the wrong thing.
That liability question has no real answer yet. Jodie Kelley, CEO of the Electronic Transactions Association, told the House Financial Services Committee in January that "many existing principles, including authorization, consent, liability, auditability, apply in the agentic context." Which is another way of saying regulators are still using the old rulebook, and the old rulebook was not written for software that goes shopping.
Forestell sees the gap clearly. At the Visa Payments Forum he laid out the hypothetical where an agent runs a thousand small purchases for you over time and eventually asks, "Do you want me to just not check?" That moment, when the AI starts asking permission to stop asking permission, is roughly where most consumers say no.
There's a deeper signal in the friction data too. A Thales survey found that 69% of consumers say multi-factor authentication actually increases their trust in a service, and 45% explicitly prefer stronger security checks even if it slows them down. The friction that payment companies have spent two decades trying to remove is the thing users now want back the moment an AI is the one holding the card.
The clean version of agentic commerce was supposed to be invisible. Your AI handles the boring stuff, the payment clears, you never think about it. The version we are actually going to get will be loud, full of verification prompts, spending caps, and a slow legal fight over who eats the loss when an agent makes a bad call. Visa and Mastercard built the rails this quarter. Whoever ends up writing the liability rulebook will decide whether anyone ever actually rides them.
