Remy Blaire: While consumers are increasingly handing their wallets to AI agents to shop and pay on their behalf, the global payments rulebook has never been written for machine checkouts. When an AI assistant gets duped by a counterfeit storefront or makes an unauthorized purchase, consumers dispute the charge. Merchants absorb the losses, and the AI developers may walk away clean. But with global card networks still relying on legacy chargeback rules built for human clicks, financial infrastructure is racing to catch up to autonomous commerce. Well, joining us live here at the New York Stock Exchange to break down who pays when an AI agent gets scammed is Pascal Podvin, co-founder and CRO of nSure.ai. Well, great to have you here. Thank you so much for joining me.
Pascal Podvin: Thank you very much for having me. It’s a pleasure.
Remy Blaire: Well, we know that artificial intelligence has been a technology and investment theme, as well as something that all of us are glued to. But when it comes to the rise of agentic AI, especially in the space of payments, who bears the responsibility?
Pascal Podvin: I think that’s one of the key questions. Obviously, when the store becomes the AI platform, when checkout itself is in the hands of the wallet providers or potentially the schemes, by the way, there is a battle between the two and we don’t know yet. To be honest, there are more questions than we have answers. But the question of liability is obviously a major one, and not only with agentic e-commerce. I think we need to talk about the rise of AI in general. But when it comes to agentic, indeed, merchants are resisting. They don’t own the internet anymore. They don’t own the store. So why would they carry the responsibility when something goes wrong between the human and the agent? The point is, I think the market is in higher demand of underwriting, and that’s exactly what we do. We underwrite the risk that is inherent to any money movement. And in this particular case, of course.
Remy Blaire: Yeah. And I think you brought up a very important point, because with artificial intelligence comes plenty of opportunity, but at the same time, risks. So give us an overview of these risks that you’re paying attention to.
Pascal Podvin: There are many ways to skin that cat, but I would say there are two major issues. The first one is the fact that the rise of AI has made social engineering extremely cost effective. In fact, it costs almost nothing. So you used to have humans call other humans and try to convince them to send money to their account. [unclear: “An agent. Now, as a productivity that’s probably 10 million higher, ten times higher than a human.”] And so what that gave birth to is a completely different type of fraud, which we call authorized fraud. It used to be that [unclear: “somebody telling you payment method”]. They make a purchase, you don’t authorize that fraud. But if somebody convinces you to send money to their account, this is authorized fraud. And obviously, it shifts completely the way to prevent that. In the past, you know, we used to focus on authenticating digital identities. So it used to be the game was, okay, if we are convinced that Pascal is the person [unclear: “really pretending to be Pascal online”], then we let the transaction in. If we’re talking about authorized fraud, that box is going to be checked all the time. And so now we’re shifting our efforts to understanding, decoding the intent.
Remy Blaire: Yeah. And you also mentioned a lot of key points there. When you mentioned social engineering, one thing that I thought about is the recent Binance hack, because not only was social engineering involved, but we’re talking about crypto here and also the emergence of national security, nation states, and the concerns regarding espionage at a time when AI and fintech are converging. So give us your perspective when it comes to that space.
Pascal Podvin: I think you have to look at both sides of the coin. Obviously, social engineering has become easier with the rise of AI. It’s become a lot easier to spin models in a short period of time, models with extreme power. But it’s true on both sides. So it’s kind of an arms race, and I’m not sure if the landscape has changed in favor of one or the other. It’s the same type of risks, the same type of fraud in different costumes, if you will. And my business is not so much in cybersecurity. We are really preventing and underwriting the risk inherent to money movement, money in, money out, across all payment rails, and we underwrite that risk. The cybersecurity side of things is very different. You know, it’s protecting against intrusions, which is exactly what happened at Binance.
Remy Blaire: Yeah. So now that we’ve covered that, I do want to get your perspective on some of the chargeback rules, especially when we’re talking about the traditional card networks or payment networks. So give us your take on that. And what do you think will happen, especially as we’re talking about the evolution of agentic shopping?
Pascal Podvin: So first of all, I believe that there will be new chargeback rules that will come up. I mean, if there is a misunderstanding between the human and the agent and a dispute is filed, that gives birth to a chargeback. There’s a new rule here that needs to be put together. And in addition to this, I think the shift in the industry and the rise of stablecoins as the infrastructure of choice, where today the norm is becoming programmable money with instant settlement, I think that also gives birth to the rise of many different payment methods. And we see that reversible payment methods, such as cards and wallets, now are being challenged with other types of payment methods that are growing, such as A2A and bank transfers, so ACH, SEPA and so on. A2A, like Pix. We see the massive rise of Pix, and we see that growing massively as well. The rules are different. We’re not talking about chargebacks. There’s no reversibility in some of those payment methods, but we still need to underwrite the risk.
Remy Blaire: And finally, before I let you go, we’re heading into a new year. And we know that a lot of the evolution in the space is gaining traction rapidly. So what trends are you paying attention to?
Pascal Podvin: So we’re obviously paying attention to the trends that are affecting AI. And, you know, we see a lot of different things happening. We’re working very actively right now, for instance, on trying to decode, decrypt the syntax of fraud, understand the grammar of fraud, which allows us then to embrace much, much larger volumes and increase complexity without having to code any of that stuff. But we can see that there are new models that are coming up. I mean, [unclear: “Jeff”] is, for instance, being talked about massively right now. The thing is, the speed at which all this is evolving is phenomenal. And there are more questions than answers.
Remy Blaire: Well, Pascal, thank you so much for sharing all of your insights today as well as your perspective.
Pascal Podvin: Thank you so much. My pleasure.
