Monica Eaton, Founder and CEO, Chargebacks911:
“Giving AI agents a verifiable identity and controlled access to money is an important step towards making agentic commerce viable at scale. Merchants need to know who or what they are transacting with and consumers need confidence that an agent cannot simply spend without limits. Cloudflare’s approach begins to address both sides of that equation.”
“But identity and authorisation only solve part of the trust problem. The much harder question comes after the transaction: what happens when the consumer says the agent made a purchase they didn’t want?”
“An agent could be completely legitimate, properly authenticated and operating within its delegated spending limits, yet still buy the wrong product, misunderstand an instruction or make a decision the consumer later disputes. In that situation, proving that an authorised AI agent made the payment does not necessarily prove that the consumer intended that specific purchase.”
“That distinction is going to become increasingly important for merchants, issuers and payment providers. Today’s dispute systems were largely designed around a human making a purchase. Agentic commerce introduces another decision-maker into that relationship and businesses will need a clear record of what the consumer instructed, what authority the agent had, what decision it made and what ultimately happened at checkout.”
“AI and machine learning will therefore have a role on both sides of this equation. As autonomous systems generate more transactions, merchants will need equally sophisticated technology to analyse those transactions, identify unusual patterns and assemble the evidence needed when purchases are challenged. Without that visibility, agentic commerce could create an entirely new category of disputes that merchants struggle to understand, let alone contest.”
“Cloudflare’s announcement shows how quickly the infrastructure for autonomous commerce is developing. The next challenge is ensuring the dispute and accountability infrastructure develops alongside it. Giving an AI agent a wallet is one thing. Establishing who is responsible when that agent makes the wrong purchase is where the real test of trust begins.”
