---
title: Friendly Fraud in Agentic Commerce: Why Every Buyer Gets a New Excuse
source_page: https://www.sardine.ai/media/the-saturday-fraud-strategist/episodes/friendly-fraud-in-agentic-commerce
canonical: https://www.sardine.ai/media/the-saturday-fraud-strategist/episodes/friendly-fraud-in-agentic-commerce
format: text/markdown
date: 2026-11-16T04:01:00.000Z
description: Friendly fraud in agentic commerce raises hard questions. Who is liable for AI agent purchases, how 3D Secure fits, and why chargebacks were built for humans…
---

**Quick links:** [Human page](https://www.sardine.ai/media/the-saturday-fraud-strategist/episodes/friendly-fraud-in-agentic-commerce) · [Home](https://www.sardine.ai) · [Customers](https://www.sardine.ai/customers) · [Blog](https://www.sardine.ai/blog) · [Demo](https://www.sardine.ai/demo)

---

# Friendly Fraud in Agentic Commerce: Why Every Buyer Gets a New Excuse

**Published:** 2026-11-16T04:01:00.000Z

Friendly fraud in agentic commerce raises hard questions. Who is liable for AI agent purchases, how 3D Secure fits, and why chargebacks were built for humans…

Every conversation about agentic commerce ends up at the same question. When an AI agent buys the wrong thing, who pays? I looked into this a while back, and I still don’t have a clean answer. When I landed at Money 20/20 Vegas, I grabbed someone who sees the answer play out in disputes every day.
Ben Herut is VP of Payments at Chargeflow. We kept this one short. We talk about who is liable, what issuers and merchants need to know when an agent is shopping, why 3D Secure struggles without a human in the loop, and how friendly fraud in agentic commerce could give shoppers a brand new excuse.
What you’ll hear in this episode:
Who is liable for AI agent purchases, and why Visa, Mastercard, and AmEx have not settled it yet
Why the AI provider is a new actor in the payment chain, and how BNPL offers a useful liability comparison Why issuers need agentic transaction flagging, and what the payment mandate, card mandate, and intention mandate could provide Why AI agent registration for issuers and merchants matters, and what merchant storefront readiness looks like
Why 3D Secure assumes a human in the loop, and what happens when an assistant buys without alerting its owner
How first party fraud grows when a shopper can say the AI agent made the mistake
Why AI provider security incentives matter, and who is most likely to pay while they stay unclear
You should listen to this episode if you:
Run chargebacks or disputes at a merchant, acquirer, or issuer and want to prepare for agentic commerce
Own authentication or 3D Secure flows and are wondering how they hold up when an agent shops
Want to understand where liability may land when an AI agent makes a purchase
Are planning how to prove what a shopper approved when an agent acted for them
### Episode notes & key takeaways
Nobody knows who is liable yet
Ben says the biggest open question is who is liable. Nobody knows yet. Visa, Mastercard, and AmEx are working toward answers, though he describes it as a step back, because a wild card now sits in the payment chain. I added that the AI provider is effectively a new actor in that chain, and I asked why only the merchant or the issuer should carry the risk. Ben compares it to BNPL. The BNPL companies were never the liable party, and liability shifted based on the contracts that were signed. He doesn’t expect AI providers to be liable, and from his experience, he doesn’t expect issuers to be either. It depends on the data that moves through the system. Until the agentic commerce liability shift is settled, that uncertainty belongs to everyone in the chain.
Issuers and merchants need to see the agent
Earlier this year, the card schemes seemed hesitant to tell issuers whether a transaction was agentic. They have since walked that back, though it is not clear how or when the data will be shared, or what issuers can do with it. Ben points to three mandates in agentic payments, which are the payment mandate, the card mandate, and the intention mandate. Without that information, it is hard for anyone to make a decision. He believes both sides need to see the agent. The issuer should know that a transaction is agentic and be a little more careful. The merchant should be able to show it is dealing with an agent, and its storefront should be easy for AI to scan. The agent has to be visible before the rules can work.
3D Secure and a dispute system built for humans
3D Secure assumes a human is placing the transaction. An assistant might buy without alerting anyone, and the owner may be at the dentist, on a plane, or asleep. In my mind, that breaks it. Ben took it a step further. The whole chargeback and dispute ecosystem was built on human behavior, and none of it was ready for a new actor. He points to the move from PSD2 to PSD3 as an example of regulation catching up, and he says it always lags. For now, there is no single answer for authentication for AI agent transactions. It changes almost weekly. We should expect the rules to move for a while.
Friendly fraud gets a new excuse
First party fraud is where agentic commerce chargebacks get uncomfortable. Ben says about 70% of chargebacks are friendly fraud or invalid, split between real unknown disputes and intentional ones. Now add a shopper who says they hoped they gave the agent the right prompt. I joked that it is the ultimate excuse for buyer’s remorse. It was a hallucination. I once gave the example of asking an agent to buy an iPhone and watching it buy a picture of an iPhone. Ben says agentic purchases are still early, and AI agent purchase disputes will reach everyone at some point. I also noted that Amazon doesn’t let AI assistants shop. Ben’s answer was that maybe they know something we don’t.
Who pays is who invests
If AI providers have no reason to block account takeover, stolen identities, or stolen API credentials, we have introduced a player that mainly wants to make money. I don’t see that ending well with today’s setup. The schemes are taking a step back to watch, and someone will pay the price. Ben agrees that it will probably be the merchants. Until incentives line up, every team should plan for it.
### Final takeaway
Friendly fraud in agentic commerce does not create new fraud as much as it changes who can be blamed. Friendly fraud already makes up most chargebacks, and an AI agent gives a shopper a very easy sentence to say. The system that settles those disputes was built for human behavior. Liability, authentication, and data sharing are all still open questions. The teams that prepare early will not wait for the schemes to answer them. They will decide how they will prove what a shopper authorized and what an agent was allowed to do.
Resources & links
Not ready to stop the conversation about my, and hopefully your, favorite subject? Subscribe to The Saturday Fraud Strategist newsletter.
Connect with Ben Herut | LinkedIn
Chief of Staff at Chargeflow
Columnist at FraudBeat
Fraud Committee Member, Merchant Risk Council
SQL, Reg E, 3D Secure, Chargeback Claims and Resolution
Connect with Chen Zamir | LinkedIn
Host of The Saturday Fraud Strategist
Helping fintechs build smarter fraud defenses
Co-author of “The Fraud Fighter’s AI Playbook”
