# Monflo | AI Agents in Smart Contracts: Can AI go to jail? - Steef Peters | ETHDam III - 2025 - Channel: [CryptoCanal](https://streameth.org/cryptocanal) - Date: 2025-10-07 - Duration: 13:51 - Topics: Education - Watch: https://streameth.org/watch/yt-drsDZaD4Cpc - YouTube: https://www.youtube.com/watch?v=drsDZaD4Cpc ## Description Welcome to the 3rd Edition of ETHDam, hosted May 9–11, 2025 in Amsterdam. This year, we brought together the brightest minds in privacy, security, and AI for a unique 48-hour hackathon + conference combo. 🌷 https://www.ethdam.com// 🌷 ------------------ Monflo | AI Agents in Smart Contracts: Can AI go to jail? - Steef Peters | ETHDam III - 2025 🎀 About the Speaker: After a study mathematics and physics at the University of Amsterdam and a PhD in elementary particle physics he made a new step designing and implementation of new technology like electronic payment systems and artificial intelligence in the financial sector. This resulted in a professorship at the Free University of Amsterdam, visiting researcher position at the Henley Business School and visiting professor position at the Beijing Institute of Technology. Next to this he started companies in the Netherlands and the Philippines directed to the design and implementation of decentral personal information systems. To apply this knowledge he joined a company in the social domain as director resulting in ways of scoring, which is now being developed in a new company, wallet79 in the Netherlands and wallet79Cambodia. He was teacher at the Hague University of Applied Sciences in the field of risk management, cybersecurity, blockchain and GDPR before staring a new company Monflo, bridging the traditional fiat currency world to the digital currency world. At the moment is is the Chief Risk Officer at Monflo and CEO at wallet79 in the Netherlands. 𝕏 Follow: https://www.linkedin.com/in/steef-peters-7928487/ https://www.monflo.com/ https://x.com/MonfloFinance ------------------ About ETHDam & CryptoCanal ETHDam is powered by CryptoCanal, an education and events platform rooted in Amsterdam, expanding into Rotterdam and ZΓΌrich. Keep up with us to see updates on future events: https://www.cryptocanal.org/ Follow CryptoCanal on X: https://twitter.com/CryptoCanal Join CryptoCanal TG Community: https://t.me/CryptoCanalCommunity Join CryptoCanal Discord: https://discord.com/invite/XJVjpCqQBz CryptoCanal unites crypto enthusiasts committed to making a positive impact. Unapologetically political, we prioritize education, events, and services while championing cypherpunk values like privacy, sovereignty, and censorship resistance. ------------------ πŸŽ₯ Credits: Intro / outro by babyPRO - https://babypro.art/ ETHDam Photography by Paulus – https://concretestate.eu/ MC of ETHDam - Collins Ejiofor - developer working at the intersection of Web3 and AI. Check out his LinkedIn for more about his work and upcoming projects! https://www.linkedin.com/in/collins-ejiofor/ ------------------ Special thanks to our partners who made ETHDam possible: 🌹 Hackathon – Bouquet: Oasis Network https://oasisprotocol.org/ 🌷 Hackathon – Petal: Circles https://aboutcircles.com πŸ’› Conference – Gold: Zano https://zano.org/ Dash https://www.dash.org/ Bitvavo https://bitvavo.com/en 🩢 Conference – Silver: Igra Labs https://igralabs.com/hero πŸ’› Conference – Copper: Lido https://lido.fi/ DeTrip https://detrip.travel/ Cake Wallet https://cakewallet.com/ The Grid https://thegrid.id/ Calimero Network https://calimero.network/ 0xbow https://0xbow.io/ Mina https://minaprotocol.com/ JobStash https://jobstash.xyz/ Cyber Capital https://www.cyber.capital/ POAP https://poap.xyz/ Acronym Foundation (Supported our Top 10 Hackers) https://acronymfoundation.org/ 🌱 Sponsor: EF Ecosystem Support Program https://esp.ethereum.foundation ------------------ 0:00 Intro 0:55 Can AI Go to Jail? 1:35 From Web3 to Web5 2:35 AI on Smart Devices 3:30 AI in Smart Contracts 4:20 Legal Complexity & Compliance 5:20 Protecting Individuals with SSI 6:30 Case Study: Cambodia Credit System 7:50 Smart Contracts vs Legal Contracts 9:05 Risks of Self-Learning Agents 11:00 No AI Jail Yet – The Legal Gap ## Transcript Welcome to East to Ethan to Ethan to Ethan to Ethan. Hopefully I have to agree with the previous I have to agree with the previous speaker that AI will come into everything. But I also have to tell you a little bit uh the whole idea is already quite old. I started with AI in 94 in a car leasing company deciding whether you need maintenance on a car or not and whether uh it was really necessary considering the payment for it. I will not go today in technology. I love it but I don't do it myself anymore. I've reached an age where I said okay I leave it to others. But I want to talk to you about this question. Can AI go to jail? Do you have any idea whether there's already a legal structure that you can send an agent to the gale to the jail? Apparently not. No. No. Not yet. Not yet. Well, it's going to happen because there is a huge hoe. Uh you can do if I'm not mistaken. Try this one. I don't know whether it's connected. Yeah, I think because it's not standard. No, I I will use the this one. Yeah, no problem. Okay, that's the question. Let me first go with a few starting points. The first thing is that I'm not talking about web 3 anymore. For me, we have web 3 developed in Monflow and Wallet 79 in the companies I work for. Uh, so that works. It's operational. It's being sold. So, that's no problem. I'm now looking at web four or five. And what is the nice thing about the definitions in web four and web five is that you will have AI agents on intelligent devices, your own telephone, right? There will be NPUs in your telephone which will be so powerful that it can certainly run AI on it. So don't believe all these technicians who are saying that your telephone is not strong enough. I had that same story in 91 when the old Motorola 9750. They said no no you cannot run anything on it. Well I said please wait a few years it will come. The same thing will happen with your smartphone. So there will be self-learning AI running in your smartphone and you will have self-learning AI agents in smart contracts. What does that mean? That means that when you make a mortgage with a bank that this self-learning AI system will check the differences in circumstances which could occur and will take action on it immediately. It will use your own personal profile at the very beginning and it will use some identification for that profile such that it knows when to increase your percentage or decrease your percentage or kick you out of the bank or anything else that's going to happen. This whole self-arning is using personal information. So you have to deal with GDPR but it will also use public information and national laws. And I don't know whether you have been at the talk of Cherry Bod but democracy is giving us a huge problem there because nations are really making their own laws. We are working in different countries. I'm used to work internationally and in a company I worked until 98, we had 21 different companies and we had 21 different system environments just to fulfill the national laws. Now this is a crazy situation but it hasn't been changed yet. So but you have to comply to it. In Europe there will be a digital act. you know that they already sued Facebook and the others. It will apply to everybody who starts using AI. If you use an AI algorithm, you have to get it certified. Otherwise, you're not allowed to use it. That's coming out of this system which we have. I can I can't help it. The other thing which is happening is that we get more and more international trade regulations. Whether it's tariffs or anything else, you will have them. So the global economy which is independent of the nation state but it will also have a lot of regulations norms rules which you have to be aware of in your software when you start using it for self-arning AI environments right now. How can we protect the individual? To my knowledge the only way to do that is using a self-suff identity. uh we had a discussion uh on on the panel that people said well no no you have to do it centrally. I hate any central system even a blockchain. As soon people start to put meta information in a blockchain I kick them out in my company. Even if they think about it, they can leave because I hate that in a blockchain there should only be one number which is an identification and that's all all the other information about people or whatever should be separated and it should to my opinion based on your own smartphone and nowhere else. Okay. So the self-learning AI for your own profile, what happens in there should also run in your phone. It should control, you should have no communication with the outside world until a certain situation is reached that you are willing to give that information to the other party. Example, if a bank want we are building that now in Cambodia. In Cambodia, the central bank has made a blockchain system for interbanking transactions. So from bank A to bank B, when you do a payment, it goes through the blockchain. Great. Then the bank said, "Yeah, but loans are still our own area." And the central bank said, "Yeah, but you are discriminating because you are using your own credit scoring." So we made a deal now with the central bank and a few banks already in Kombucha that we will build a credit scoring system on your phone using the blockchain information and your personal information but we will do all the calculation on your phone. that will end up with a credit scoring which will be equal for all the banks to be used and you can know in advance when you for example want to have a loan of $10,000 you know in advance whether you can get it or not and you select for yourself do I want to have it with bank A, B, C or D. So you are in control of your own decision regarding your loan. You determine whether you ask for that loan instead of going to the bank and said I want to have a loan. Oh, then we want you to have your salary slip. Do you have a mortgage? Do you have this? Do you have that? And all that information is stored with bank A, bank B, bank C, bank D. I don't want that. Now, we have one I have a proposal here about self-learning in smart contracts. But then we get into a problem. How do we know? I think I skipped one, didn't I? Yeah. So, this I explained, right? You have everything in your smartphone, but now we get a problem here. When you have an AI system as an agent running in a smart contract, I I hope everybody knows that a smart contract always executes, right? You cannot stop it. A legal system, I'm from the Netherlands, from Europe. A legal system in the Netherlands is normally that you make a contract in advance and when things go wrong, you go to the judge, right? So it's in fact always afterwards expost. Do they call that the expost situation is the normal legal situation that you use a normal written contract. In case of a smart contract, it's a completely different situation because a smart contract is always made X under. you fully write down what has to be executed when a certain circumstance occurs. You cannot stop it. Now assume that you have some self-learning AI agent in there which means that it could find some data and says just like in the movie I robot I decide all these humans I should kill them because they are ruining the world right I'm for environmental circumstances so I'm going to kill all humans now I think he's correct but it's not according to the law yeah you agree but it is in a smart contract so it will be executed that's was perfectly shown in that movie I robot it will just continue killing if you don't kill the robot so that will happen here as well governments are only looking x they only look whether you make the algorithm and the agent in the correct way but they are not involved in the execution so they might go something else here is the problem where AI has to go to jail when he goes when he does something wrong. When do you go to jail? That is when you are acting not according to the laws and somebody finds it out afterwards. But how do we find it out with an AI agent? The only way you can find out is afterwards by having an incident management. And that's what we do as bankers. uh my background is also banking and yes uh we have n I'm a risk officer so I know exactly what to do with incident but that's afterwards then it already happened how can I detect when there is a cyber attack which changes my AI agent in the smart contract for example I loved the previous presentation because the guy said that agents can be very eyes to be used because they can detect a lot of things but I can also as a hacker I can change those agents how can we handle that in fact that's the big problem which laws do we have to comply to what are those laws and how are they used and how are we going to detect the cyber attacks who are making changes in the AI algorithms At the moment, governments are only looking at compliance for AI agents for XU under. So, you have to get them certified before you are allowed to use them. There are no regulations yet regarding when AI does something wrong. So my advice to everybody is when you make AI agents who are self-arning, there should always be some possibility that it will report what is happening. What kind of decision is being reached before it makes the execution otherwise it won't work. on your telephone is not a problem because there you will be s in control yourself. You will see the result and you know whether you want to act on it or not. But it's a different story when I make a smart contract in a blockchain and the smart contract contains the AI agent because then it's running on the blockchain immediately and then I have to have other measures to be able to detect that something goes wrong. That's in fact my story. I know it's theoretical. I'm sorry about that. I'm not going into the technical and programming, but I hope I've given you some ideas how to apply AI agents in your own smart contract environments. Thank you. Thank you very much. Great.