The Payments Experts Podcast
Expert payments attorneys discuss the electronic payments industry from a legal perspective.
The Payments Experts Podcast
AI Hallucinates a Refund Policy (It Didn't Exist) & The Business Paid: Rogue AI in Payments | PEP105
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Software that negotiates prices and completes payments for you sounds convenient until it hallucinates a refund policy.
Software that can shop, negotiate, and pay for you is no longer science fiction, and it is already colliding with the realities of payments risk. We sit down with Dale Laszig from the Green Sheet (https://www.greensheet.com/) to break down agentic commerce in plain English and explain what changes when “the customer” is a digital agent acting autonomously at real-time speed.
We unpack agentic commerce, where software acts on a buyer’s behalf and can search, negotiate, and complete payments without a human in the loop. We connect the promise of automation to real risks like hallucinated refund policies, AI-driven fraud, and the need for tighter contracts plus continuous monitoring.
• Defining agentic commerce in plain English for payments teams
• Why rules-based AI can be safer than LLMs
• The airline refund story and what it teaches about liability
• How AI changes chargebacks and dispute response workflows
• Deepfakes and synthetic merchants targeting onboarding gaps
• The shift from one-time KYC to continuous behavioral monitoring
• AI versus AI dynamics in fraud and risk decisioning
We dig into a memorable cautionary tale where an AI system hallucinated a refund policy and the business had to honor it, then connect that lesson to chargebacks, dispute management, and the legal pressure points that show up when machines make commitments. From our perspective as payments-focused counsel, the practical starting point is updating contracts, policies, and training so liability is clear and teams know how to respond when automation goes sideways.
From there, we get concrete about the fraud landscape: deepfakes, synthetic merchants, fake documents, and the growing gap between merchant onboarding and ongoing behavior monitoring. The big takeaway is that “set it and forget it” KYC does not hold up in an always-on world. We talk about building a multi-layered trust infrastructure with strong identity signals, behavioral monitoring, governance frameworks, and AI-powered fraud detection tools, because it often takes AI to spot AI.
AI fighting chargebacks meets AI pushing fraud. Who wins when machines argue at scale? We talk contracts, liability shifts, and why you should partner with security experts instead of building tools yourself.
If you work in payments, underwriting, risk, or compliance, this conversation will help you think clearly about agentic commerce, AI fraud, and what readiness should look like right now. Subscribe, share this with a colleague in the industry, and leave a review with your biggest question about AI in payments.
**Matters discussed are all opinions and do not constitute legal advice. All events or likeness to real people and events is a coincidence.**
PEP Links:
https://www.globallegallawfirm.com/podcasts/
https://www.buzzsprout.com/2176695
A payments podcast of Global Legal Law Firm
Get Expert Help Early
SPEAKER_00Have somebody a professional review your contract before you sign anything. And I would say in the same tone, you know, go to a security expert that's been doing this, that's, you know, living, breathing, eating, sleeping this technology. Don't try to build it yourself. It's it it would take way too long. It would cost too much. And how can you be sure that you're going to cover all of the bases? You'd you're better off going to an expert. And there are a lot of really good ones out there. I really hesitate. I mean, I'm I'm talking to everybody, so uh I'll keep that part as deep background.
SPEAKER_02All right, perfect. So maybe people can contact you for recommendations.
SPEAKER_00Okay, definitely.
SPEAKER_02They'll direct message you.
Welcome And Guest Introduction
SPEAKER_01Welcome to the Payments Experts Podcast. A podcast of Global Legal Law Firm. We hope you enjoyed this episode. Welcome to the Payments Experts Podcast, a podcast of Global Legal Law Firm. We're really excited. We got uh joining us in studio today, managing partner of Global Legal Law Firm, James Hucker, as well as our special guest, Dale Lastig, with the Green Sheet. Dale is the senior staff writer over you can find them at thegreensheet.com. Dale, real pleasure having you on the podcast. Welcome.
SPEAKER_00Thank you so much. It's a pleasure to be here.
SPEAKER_02Yeah, Dale, I'm excited to have you on. I think I've told you before, I think it's actually pretty well known. The Green Sheet is the best payments periodical out there. You focus, I'll say, and maybe this isn't true, but I feel like you're focusing mostly on agents in the ISOs, and they're the ones that need the information to operate in this industry and you get it in a nice, you know, concise packet, where some of the other periodicals is, you know, it's really broad, but I feel like the green sheets, if you're in this space and you're an agent, and I know everyone I talk to, they all read it. It's the one to read.
SPEAKER_00Thank you. That really makes me feel good. And to your point, we did begin our life as a uh periodical, and uh, and we grew into a really robust online resource, but our original readers and uh very much our core audience are still the people on the front lines who are making it happen, moving our industry forward. Uh, we used to say one merchant at a time, but it's not that way anymore. Everybody is scaling up and it's enterprise level at this point.
SPEAKER_02Yeah, it is. It's uh yeah, it's a different ball game.
Agentic Commerce Defined Simply
SPEAKER_02So let's jump right into why you're here. You wrote two articles for the green sheets on agentic commerce. And I will admit I was totally naive to this being out there. Um, not at the time of your articles, but it's only recently come on my radar that this is happening. I think it's because I live in a bubble and I just don't really buy things. But uh tell us in plain English, uh, what what is, you know, for the ISOs, the ISEs, and cars, what is what is agenc commerce? What's your, you know, Webster's definition?
SPEAKER_00Okay, so um first of all, thanks for having me. It's a pleasure to be here. And uh and I'm also happy to mention that um my title has changed. I'm now content strategy director at the Green Sheet. All right. Um, but uh anyway, I think the easiest way to explain agentic commerce is when software stops being just a tool and it starts acting on your behalf. And this could mean going out, comparing products, evaluating, negotiating a price point, and completing a payment all on your behalf.
The AI Refund Hallucination Case
SPEAKER_02That's right. And then on the, you know, I'd say it on the other end, you know, one of my favorite parts of your article that I loved was the company. I can't remember the company off the top of my head, but their their robot promised a refund to somebody. And they had to honor it.
SPEAKER_00Oh, yes. Oh, that was uh that was a fascinating uh case study, yes. And I mean what was an airline.
SPEAKER_02Yeah, go ahead. Sorry. Yeah, why don't you tell us a little bit about that? Because I but that one I loved.
SPEAKER_00Well, it was a fascinating case study of uh an airline, actually, it was a large language model, and some of the experts I've interviewed encourage everyone to really stick to rules-based AI because LLMs have been known to hallucinate. And uh in this case, uh the uh the AI made up a policy that really wasn't there, and uh they they granted a refund uh to a customer that really was not entitled to one. And uh the upshot was the airline had to honor that. Uh so um that was uh like what we call a terrible warning in the industry. If you can't be a good example, at least be a terrible warning.
SPEAKER_02At least right. At least be show everybody what to what to worry about. Because yeah, I mean taking chargebacks to another level. Yeah, yeah.
SPEAKER_00And that that's very much in your realm. I had I have to imagine this is this must be a very exciting time to to have a law practice and you know to be so focused on ISOs and agents and ISVs. Um I'm just fascinated to know more from your perspective how that looks. Um the legal frameworks, are they flexible and resilient enough to support these incoming technologies, or will will you have to really um write new legislation to uh to protect our industry from some of the uh the very real risks involved?
Contracts And Liability In AI Payments
SPEAKER_02Well, there's almost no legislation that protects this industry from, you know, there's almost no legislation that even touches this industry, actually. So I don't see that happening, anything targeting uh the payment space as related to protecting this. But yes, from the legal perspective, we're all over people about saying you absolutely need to be updating your policies and procedures because this is out there, it's happening. And we've had one, you know, one client bring on a situation where the chargeback, they have their AI fighting chargebacks and responding to it. And they're going, now we realize it's AI trying to push the chargebacks through. So the machines are literally fighting each other here. And whoever's technology is gonna win, you know, in this case, you know, somebody got really creative with the airline to go, yeah, absolutely, you know, forget the the chargeback. You're giving me a refund because it's your policy that I wrote and doesn't exist. So, you know, as far as, you know, what we're doing is it's just putting it in the contract and trying to shift liability around because we know as the ISO or the ISV or the agent, it all flows down to them. And they're ultimately the ones that are going to be holding the bags. So putting language in your agreements to protect yourselves and then training your employees, creating training material for people on these things because the attacks are happening. And you described a couple of them, you know, deep fakes and synthetic merchants. They're creating fake websites, Facebook pages, documents, and it's not that hard. We use uh an AI in our software that are in our firm that just takes control of my computer and it will go do things, and I'll leave my desk and I'll come back and I'll be hey, you know, hey, I'm all done. And I don't see why it couldn't do you know everything that you list of creating a fake, probably fake business. I think it could probably even go on and form a business, you know, if I gave it my credit card information, so uh, which I won't do, but um, yeah, I mean, what do you see as the you know, the weakest KYC and underwriting controls related to these kind of deep fakes and synthetic merchants?
Deepfakes And Synthetic Merchant Risk
SPEAKER_00Okay, that's a great question. And I guess um what I'm seeing is uh, and and what the experts I've interviewed are seeing is uh there's a gap between onboarding and constantly monitoring behavior. Uh in the old days, in the old paradigm, we could just unboard a merchant and then move on.
SPEAKER_02Right.
SPEAKER_00Uh and we can't do that anymore. And I think the main reason we can't do it is because everything is real time now in our industry and everything is happening very rapidly, the technologies are
Continuous Monitoring Beats One Time KYC
SPEAKER_00agile, they're autonomous, they're always on, always connected, and you just can't you can't have any one and duns anymore. You know, whether it comes to onboarding a merchant or or just um verifying someone at the front door before you let them in. You can't just say, okay, they're verified. You have to continuously monitor that behavior to really make sure that they are acting um on behalf of someone that they they are you know, they're mandated by a real human uh if it is a digital agent and not a human. I mean, look at look at all the years that we spent with KYC, and I'm sure they kept you very busy too in your role. And now we have to do it all over again. Right. Uh only this time around, it's not with humans, it's with digital agents. So it's gonna be a fascinating time and uh a very good time to be an attorney, I would think.
SPEAKER_02Well, we hope so for the good reasons. I mean, I I like what you said of you know what you can't set it and forget it, but that's been the rule for a while. No one on you know, e-commerce, you know, really started booming, and then fraudulent, you know, players figured this out. We would have clients that, you know, if you didn't catch this thing, they can crank you know 700k in a few hours, you know, until people started building the tools to shut everything down and all of the you know, everything to fight chargebacks and fraudulent uh merchants. But I see this, yes, of being, you know, the second coming where the current tools, they don't, they're not gonna, you're gonna be able to get around these. And now when you have people with, you know, they have their 25 merchant accounts that are all you know potentially selling the same or similar products. When you put it on that scale, the current tools, it you'll be able to trick them. And people will figure that out. So, you know, right now I think you know, people going, well, we're using AI, like we said, it's gonna be AI fighting AI. And usually in the beginning, the criminals are smarter until we see, you know, how smart they are, then you can build the guardrails around it. But the problem with this is that it just keeps going and getting better by you know the second, right? So Right.
SPEAKER_00I mean, we're talking about self-learning technology.
SPEAKER_02Right.
SPEAKER_00So uh pretty hard to keep up with that. So what do you I mean have you have you uh I mean you you're using an AI, right? So you uh upload a document into an AI, and uh what it it takes, what, one nanosecond for it to come back with a lengthy explanation or you know, and and catch every little typo in that document? And it's just astounding, but that gives you an idea of how fast and and the scale in which all of this is happening.
SPEAKER_02Right. So what does it look like right now?
AI Versus AI Fraud Arms Race
SPEAKER_02What does you know, readiness look like to fight this? Do you see are you do you see people putting technical protocols, putting people there? Or, you know, what are you seeing out on the street?
SPEAKER_00I I think so, yeah, definitely. The technical protocols, um, I've seen it before in our industry. They eventually become policy. Uh, they're a great way of getting everyone together. We need to be interoperable, and we need to make sure that our trust infrastructure is evolving at least as fast, if hopefully faster, than the bad guys. And I think um the trust infrastructure has to be multi-layered like everything else in security. I mean, basically, you need a strong identity infrastructure, you need behavioral monitoring, and then you need some kind of governance frameworks. I think if you have all those things, and if you're in ISO building a stack, you really need to think about having an AI-powered uh fraud detection tool in your stack because it takes an AI to recognize another AI.
SPEAKER_02Right. It can spot itself. Are there tools out there? Are there tools out there that you can recommend that people are looking at, or is this build your own?
SPEAKER_00Well, I definitely think it makes sense to partner with leaders in the space. Um I mean i i in the same way that I would I would suggest to an ISO to go and and we've been saying this for years, you know, have somebody professional review your contract before you sign anything. And I would say in the same tone, you know, go to a security expert that's been doing this, that's you know, living, breathing, eating, sleeping, this technology. Don't try to build it yourself. It's it it would take way too long, it would cost too much, and how can you be sure that you're going to cover all of the bases? You'd you're better off going to an expert. And there are a lot of really good ones out there. I really hesitate. I mean, I'm I'm
Building Trust Infrastructure That Works
SPEAKER_00talking to everybody, so uh I'll keep that part as deep background.
SPEAKER_02All right, perfect. So maybe people can contact you for recommendations.
SPEAKER_00Okay, definitely.
SPEAKER_02They'll direct message you.
SPEAKER_01Thank you for listening to this episode of the Payments Experts Podcast, a podcast of Global Legal Law Firm. Visit us online today at Global Legal Law Firm.com. Matters discussed are all opinions and do not constitute legal advice. All events or likeness to real people, and events is a coincidence.