Panel | Crypto was made for AI Agents | Common S3nse - 2026

Transcript

amount that can be taken out, I would feel much better than passing on my credit card details already. Like yes, you can set limit limits on credit cards as well, of course. But for me, it's okay if I could have an agent that can do a ton of small transactions for me just from my crypto wallet, that would save me quite a bit of like time and effort. And I realize AI is making me incredibly lazy, but just okay, I need an Uber. Like book me an Uber and pay for it. I want a pizza. Like buy me a pizza and pay for it. I think we don't yet realize in how many sort of small day-to-day things we need to pay for something. And if I can hand that off to something else to do do it for me, I would be very happy to do so.

I I guess it's the same in every field where AI starts getting used, right? At the start you're like, oh I can do some small things, but I can't really see it getting super useful. And then a year later, you've handed off everything and you're like, okay, I I barely need to look at code anymore. Um so yeah. >> Yeah. Now on that thing, um how much of that use case is AI agents setting up the sale and how much of them is transferring the money? Like if it's just here's the flight and you validate, you send crypto or whatever yourself every time. You just click, click, click. Does it does not needing to do the click, click, click click, is that save you so much time that you're like, I really need to invest in this use case?

>> No, for me I think that's fair. That would already save a ton of time if it just sets it up and I do the final click. Um I think that's kind of like where do you set the boundary, right? Maybe for all payments under $5, you're like, okay, you can do it yourself. If it's bigger than that, yes, let me do the final check and I'll do my like fingerprint to validate it. Uh yeah, that doesn't cost that much extra time. But just for a ton of small payments, yeah, could be could be useful. >> The $5 limit rule. >> Yeah, yeah, Yeah, [laughter] yeah. >> Well, I mean this is interesting, right? Because like we don't have like we have things like Amazon. I was you know working on a project that with MasterCard that wanted to connect, you know, crypto and their

own payments to like a a talking animation thing. There's all these these there's all these possibilities, right? But if we're discussing like, okay, is AI made for crypto? I mean, the answer is yes, because what is the AI should be looking at the data, right? And then reacting to the data. That's what it is. It's artificial intelligence based on past data, which you have entire purchase and trade histories living on every single chain. And that's where like the the use case comes from.

>> Yeah, and to the point of them being made for each other, I mean, blockchains. I think honestly, the key reason why they're they just introduced like a standard API layer that all these apps can talk to in a relatively secure protocol layer that they could all interact with and kind of speak the same language. So, it's like totally natural that they would all be able to plug in together and interact with each other. Um and yeah, I use agents mostly for like testing my app and building, you know, like scenarios where it you'd have people trading things and um moving funds around different smart contracts. So, it's yeah.

>> Well, this is why Nasdaq is moving to Canton, right? My my company is currently in the Canton accelerator and the the VC I'm representing for Otter Tank is running it, but this is this is exactly the reason. This is the underlying tech and how it can be used. Because once we figure out a real world use case for this technology, blockchain, you know, rather than than just throwing things at it saying, "Does this work? Does this not?" We solve a problem. Does it make it easier? And in my opinion, please feel free to disagree, but And is yes.

>> I I think we like we know the use cases, the core use cases of blockchain and then and that's proven. It's just the rest of the world hasn't caught up. To be honest, I think we all know that. Like it's obviously more efficient, permissionless, like we all love these things but like the use case for your use case, okay, maybe there's there's some portion of the retail population that does all these transactions all the time. I don't buy my stuff. So, it's like does this really help me? Am I really going to use it? Probably not. But I could see the edge cases for certain guys like you or but mainly for business. Like say when you interact with the when everything gets plugged in and you're a contractor, like trying to pay invoices for a construction project. You have all these different things. It's like here's the

budget, figure it all out and automatically every different payment stream from old school to new school, figure it out and pay it all on the crypto rails and with the API. Like that is cool, but like that won't come online today or tomorrow. And that's that's kind of what the delay with blockchain on these tags. Like we have all this stuff we talk about agenting payments. The guy outside running his business doesn't know what you're talking about. He doesn't want to take payment with your Zcash, no offense Zcash, your QR code or he's like he's slowly going to move on momentum and we'll finally get there and then maybe we'll see this super cycle start start pushing out there.

>> Yeah, and the business transformation use case is huge, but also these companies like can't hardly get insurance if they integrate these at this point just cuz it's so early, but I think it's coming. We'll see if the security rails catch up. >> Now, how much of this use case can be satisfied with current fiat rails? Like say you just have a company card, you don't have to KYC for a company card, you just set it up and you give it to an agent and set a spending limit and then when it gets close to the spending limit it just tells you, "Hey, I need more money." And then you just look into it and raise it or whatever. So, like using traditional fiat rails first, not even stable coins yet, but those. What's the big problem with that from an agentic standpoint?

>> Well, it's it's slow first off. Like any of these um, what do you call like, uh, Mercury, like any of these banks that are fintech banks, everything's operating on the the legacy rails, ACH, the big delays, all these things, and they just abstract that away, so they take the brunt of all those delays. You just don't see anything on the client side. So, they still have accounts receivable, accounts payable, you just don't see it. So, that's how it's always worked. And the same thing will go with agentic payments, you'll still have someone having the brunt of this kind of delay, moving the funds where you don't see it happening just so they can make the experience seamless for you. Uh, so I think that'll exist. I don't think it's relevant because you have that's another layer where people can make a business on is by making it more efficient. Um,

so to me that's not a barrier. You're still going to have these fiat rails as long as people are still comfortable. Uh, Visa and MasterCard still settle to the fiat system. Um, that won't happen overnight where we change it. >> I I think another thing is, maybe it's my agents reflecting on me, but they are very impatient usually. So, when they want something, they want to get it right away and as efficiently and quickly as possible. And I think, yeah, the business case I think is different, but for small things that they need to purchase like the little bit of data, for example, sometimes those payments are even too small for traditional finance, right? Like talking really micro payments. Um, also on our website, like many of the payments are literally 1 cent. And then that's a case where

it's just not possible to do with credit card. You can do it with like creating an an account on a website depositing some crypto there and getting an API key or whatever. What we see on our end is that many agents don't want to go through the effort of actually navigating to a website, creating an account, doing everything. They just They need something now, they call the API, they get a response back that says pay X amount of USDC or whatever to Y address, and they get it immediately. Um, I that's what I noticed running the agents myself as well, and I think, like, we value privacy, so we don't really know, but we think that's how people are using it on our service as well. Just What is the quickest, most efficient way to get to my target, which

is to acquire this data or to generate a video, generate an image, or anything. Um and fiat system is just not suited for those kinds of very small, very quick, very sort of loose transactions, I think. >> I would add onto that and just agree and say, yeah, it's the fragmentation of traditional finance. There's a million different APIs that you have to understand how to talk to. You have to get credentials for each one. Um at least with the blockchain you can see, you know, and get verified s- smart if it's smart contracts or whatever, you know API uh endpoints that you can pull down and just talk to, usually permissionlessly.

Sometimes permission, but depending on the app. >> Yeah, now think about stablecoins cuz this seems to be the the first way this is going to be used a lot, right? Um is there any time when that starts to become less desirable due to things like, for example, a drug sanctioned address blacklisting, things like that, or maybe it hits like the Change Lee shotgun KYC threshold where you try to swap like a few hundred dollars worth of something and then now that the agent needs to KYC and it can't and then you just have your money hostage?

>> Well, this is exactly one of We have a circuit breaker set up and this is exactly what it does. It checks the sanction wallet, it checks the degree of separation, and then the uh the person selling the items would be our client and they can stop the transaction before it happens. And that's control, right? This is It's a huge database. It's a huge set of data we're acting within to make transactions. There's a solution. >> I think the rules will change. Like first off, the BSA Act is will be shut down at some point or revised because look at the robo-taxis, right? Like what if you have your Optimus robot driving

you around and he gets pulled over. Like does he need a license? I don't know. But, this is These are all things that really >> his rights? >> Does he Does he get Miranda rights? Like, does that exist? And And laws have to adapt to it. It's just like I remember you couldn't have a screen in the front of your car back in the day. It was illegal. And now every single car has that. Like, laws need to adapt to this new tech. So, if you have stables and you you have this high-volume agentic finance, like there's not going to be blockers like that. They want especially in America, they want finance the rails to move very smoothly. So, you're not going to stop that with They'll find some way around it is what I'm saying. But, I think things need to adapt. We need to realize that KYC is

totally flawed. And hopefully we can make some some wise changes to this. >> I I also kind of wonder just in terms of This is kind of off-topic maybe, but the agents might realize like, okay, there are these limits, right? The sooner as I do 100 USDC, I start running into limits. Where can I swap this cheaply into another coin and swap it back so that I reset my limit? I wonder if if that's actually a thing because I know at one point >> bot. >> Yeah, no. I like at one point I gave it access or I gave one of my agents access to a wallet and it wanted to pay for something, but it couldn't pay in the coin that I gave it and it ended up swapping to a different coin to then pay for it. Uh which I wonder like how that

would work just from a sort of the KYC perspective, right? Like, it's very easy to swap into Monero and back. >> And I would just say um from Yeah, you can you can control things on a agent perspective where the agent is making decisions about when and how to send the funds across, but also a lot of these apps can build like, for example, vaults. Uh if you had a vault that was managed by an agent, you could have like thresholds where they could trade only within this range or whatever. And then if they go price goes below, then they're not allowed to touch it anymore. Same thing with KYC rules, you know, they can plug into a smart contract that reads a list of banned addresses or whatever. I mean, depending on your legal requirements, but yeah, that all makes it pretty easy.

>> So, do you think we're going to these kinds of pressures are going to take the stable coin world, let's say, in specific and kind of loosen the old school KYC rules around that that are very rigid and built for humans, or might it go the opposite direction a little bit where now you're starting to have more safeguards built in. Now, all of a sudden, you know, state like Tether's been freezing a lot of coins lately. And like is that going to impact the attractiveness for stable coins? Like, how often is a stable coin going to be or your agent going to be like, I need to buy these things, but I'm probably just going to swap USDT into I don't know, Nano or something else that's like also very fast, but is not doesn't have those censorship kind of things built in

and just kind of goes that way. >> Well, there's the travel rule, right? So, there is going to come a point where everybody people are going to have to prove these these transactions and where it came from, and you can't an agent take the ball and run with it. >> I think it's a losing battle for the like the old old guard. I just think technology and and the we have the best nerds on our side and privacy side. And I just think the whole thing is stupid and the majority of the population thinks it's stupid.

And I think stupid things just eventually get pushed out by tech. Like, if you have an agent that could swap through six different pairs permissionlessly to get you the outcome that you desire, and things will adapt. And and that's just going to be the case. Like, Tether could freeze all the funds at once. Well, there's a million other coins we can transact in that still have value. So, there's always a workaround, so it just makes me think that things will change for the better. I'm an optimist, though. >> I also think the freezing is always slower than sort of crypto can move, right? Like, if you have an agent that can hold its funds in, let's say, Monero, Nano, whatever, a quick coin, ideally, and it just swaps a little bit into USDC when it wants to pay for

something or like the exact amount that it needs. Yeah, sure, maybe the USDC gets frozen after, but like it's already paid for it at that point. Um so, I think honestly, I think governments and regulators will try to be more strict, probably, on this whole subject, just on transactions in general, because it's scary. But I think it's yeah, like you say, it's going to be really difficult just to to actually manage to do anything about it because it just moves so fast, right? >> Absolutely. I mean, look at look at all the hacks. Where do they go? They go to Bitcoin or Ethereum or Monero, like something where no one could touch it, no one could freeze it. And like you said, just just cash out when you want to transact.

>> Yeah, and to that point, uh I think we just saw this giant hack in the US this week where they got all the credit card information of like all these people who did KYC, basically became a giant honeypot. So, that's that's good. Um and uh yeah, there are projects, I think, that have really good approaches to managing KYC on chain, like the Oxbow project. Uh I think they use like some kind of zero-knowledge proof setup to prove that you're not you you still have that shielded pool, but you know you're not talk sending funds to someone who's sanctioned because it's split off. Um so, I'd like to see the innovation happening, for sure, even if KYC is super annoying.

>> Now, what does the agentic economy do for liability? Because uh as far as all the things agents can do, I'm sure they can do tons of sketchy things, but the the most impactful thing that they've been able to do on a on any kind of scale seems to be paid for stuff, where usually there's a lot of liability and responsibility on the individual doing it. And we haven't had like payment bots kind of thing the same way. It's usually a person behind it. And the laws are, you know, not just the restrictions, but also the legal liabilities kind of built around the assumption that will be people behind all these. And now I'm obviously, if you tell your bot, go buy, you know, a bunch of stuff on the dark web, like

okay, if that prompt gets intercepted, I'm pretty sure, you know, there will be direct legal liability. But, what happens if you have just some more open kind of loop going on, and then it just does something that ends up being illegal or some other kind of thing that you had no input into the decision, where does the liability fall on that? Well, both currently and where do you see it kind of going? >> I mean, I think we'll figure it out soon with the hugging face hack, right? Like, OpenAI did not intend to hack hugging hugging face, probably. But, the prompt they gave to one of their models ended up hacking hugging face.

Um who is liable? Like, is OpenAI going to Like, in this case, I think they talked it out with hugging face, essentially. Uh but, that is going to happen. And I think I'm not a legal expert in any way, but I think at that point, um the sort of talk was around what is the intent of your prompt? Like, it's more about the intent, and then you are less responsible for what ends up happening. But, I don't know if that's like scalable or is is that going to work out in the long run? Yeah. >> Yeah, and speaking of liability again, as a business, you just can't get insurance to put agents on your stack at this point. So, that I think says it all. Like, there's they can't even uh you know, underwrite the risk accurately yet. It may come when they can and

security is locked down enough to do it, and we hope it does. But, we'll see. >> I think liability is an easy one when money's involved. Who Who benefits? Like, who gets that? Who's got the control of the seed? And that owner of the asset will probably be held liable. >> How liable? Cuz that's that gets a little bit interesting. There's a very different thing between, again, intent to purchase something illegal versus you did everything in your best, you know, your best kind of knowledge to do something perfectly above board, and then the agent had some kind of a glitch or malfunction and did something, and I don't know, the answer seems to not be no liability, because then they would be just oh, it was a completely autonomous

agent that malfunctioned to do all this stuff. That That loophole probably won't be able to exist, but then the opposite of your your liable for everything your agent does in a very direct way probably would cause a lot of chaos as well. >> Yeah, just prompted with make me rich and then be surprised when it starts scamming people and uh >> Your honor, I said make no mistakes. >> Yeah, I know. I I I don't know. I don't think this is like it would be impressive if we we figure it out here, right? Between the five of us. >> You guys are smart.

>> And we're all lawyers, clearly. No, I I would love to have a good answer. I I've no clue how it would work. I think the right now it seems to be still the impact that these agents have in the real world is very limited, but not so much anymore lately since they've started hacking stuff. Um so yeah, you're going to probably need guardrails because I don't think there's a world in which we conclude you have zero liability for what your agent does. Like at some point you're going to trace back to okay, but what did you prompt it with? You should have expected like this to happen.

Yeah. >> I mean, no one's set the legal precedent at this point. Like think about when when the internet first appeared, right? They There weren't any laws on the books. Like it was uh you know, the FCC had all these rules if you had a spectrum and you broadcast, there was decency rules and I'm speaking from a US perspective, but you could be fined for disseminating certain information, defamation, this and that. Internet, there was no rules on the books. And it was a free-for-all up until you know, the first legal cases started happening and precedent set. So, we're at that stage with the AI agents.

I haven't heard of anyone being prosecuted or tried and I think someone's going to find out and some rules are going to be written and ultimately I think it'll it'll come back to tracing the money and if you had malicious intent, etc. etc. >> It's also worth mentioning that to shill my project a little bit, agents will have identity and when you have identity, you can have reputation. And when you have identity and reputation, people can kind of track what you're doing, um understand the risks and the past behavior behind your specific setup and have a better idea of what you're going to do going forward. So, hopefully people will adopt those kinds of market-based measures of figuring out where risk lies in the, you know,

universe of a bunch of different AI agents that are assigned to specific tasks. >> I agree. I mean, they have to. >> All right, looks like we're out of time, but thanks to everyone for being here. This is a very fascinating conversation. Hopefully everyone sends some crypto with their agent today and let some loose and let me know let me know how it turns out. Thanks everyone.