Ne faites pas confiance, vérifiez : la souveraineté à l’ère de l’IA | Bitcoin 2026

All right, welcome folks. This uh panel is don't trust verify sovereignty in the age of AI and I have a great panel lined up and maybe we'll start with a round of introductions. TC why don't you begin? >> Uh hi TC here. I created uh timechaincal.com and there's now a mobile app timechain calendar. Uh I created it. some AI was used. >> So, my name is Miland Deeda. I co-founded NanoGPT and we run a.bitcoin.com, which is essentially a way that anyone anywhere can use any AI model privately. >> Uh my name is Charles. I'm the founder of SAS for AAI, which is essentially a platform with AI tools um paid with lightning, so no sign up or anything. Um and essentially it's for humans uh but also for uh agents and we offer services as uh simply as simple as uh generating images uh videos but also more niche product like having your AI make a phone call or send a text message or even uh take an EPUB and and turn it into an audio book. >> Great. So Charles, let's stay with you. Well, everyone on this panel is building something that involves AI deeply and you're all Bitcoiners. Where does Bitcoin fit into the AI equation? >> Right. So, I've been working in AI and in bit in Bitcoin since 2017. So, in in in those days, I was very optimistic about the power of those two technologies um to make it essentially to to become freedom tech. Um, I got a little worried when chatgpt came out and I saw the adoption that, you know, everybody kind of threw themselves at this product that was not private. Um, you know, essentially collecting your data, permissioned, um, and heavily censored. Um, so I got a little worried then that this was going to be another one of those big tech monopoly where, you know, winner takes all and that would be the end of it. uh which is you know considering again the the the capacity of this technology is very orwellian. Um so but I got hopeful again uh once I saw the first kind of openw weight model come out. Um and this is when I decided that you know I had to kind of contribute to that event of of joining Bitcoin and um you know open source or openweight um models. Um, and the idea is that both of them contribute um to helping you again be permissionless um and private and and censorship resistant uh because it's a it's a weakest link logic. If you're using, you know, freedom tech AI, but you're paying for it with money that's not private, that's permission, that's trackable to you, then essentially there's no point at all. It's like building a fence around your house, but only only the left side. Like you might as well not do anything. You you need the full fence. So I think that's one of those things where when you combine the two, this is the only way that you have a tech stack that's actually really again freedom maximalizing. >> Malone. >> Yeah. No, for us the reasoning was largely the same. So AI is kind of eating the world. I think we we all agree. I think we all agree nowadays. Um but not everyone has easy access to it. And part of that is not everyone has credit cards. Some people live in countries where they wouldn't trust their government to collect data on them. Um, so they need a private way to access these models, not just openweight ones, but also realistically the best models at the moment, ChatGpt, Claude, which are closed source. So we figured, okay, what can we do to make those models accessible to anyone, give everyone access to the top models, but make it so that it's not trackable to them. So, you don't have to pay with your credit card. You don't register with an account. You can just put in a few dollars in Bitcoin, even less actually, and use any of the top models. Um, we think it's important to keep AI the top AI models available to everyone because it's a better world when everyone has access to them. So, yeah, we think a we think Bitcoin plays a super important role there. >> DC, you your approach to AI is a little different. And I think both Charles and Milan are building a suite of products that has AI deeply embedded in it. You actively use AI for everything you you or a lot of the things you create. Where do you see just the same question to you, but from your perspective, where does Bitcoin play a role in this? I I think there's a lot of people out there with really great ideas about how to communicate Bitcoin and help people with the steep learning curve and the broad depth of knowledge that there is to bring to the surface. There's there's a lot to build. There's a lot to do and most of us only have a specific specialty in potentially one aspect of things or in my case I was a software engineer for years and I did a specific role in larger software teams. Um, and so if I want to launch my own product, there's actually quite a lot of different kinds of work to do there. So I I see really AI as a powerful uh piece of the the the puzzle as far as overcoming a learning curve and overcoming new skills getting added to the the work that you're doing. anything with any degree of sophistication that's technology- based, whether you're building a Bitcoin wallet or in my case building more of a data dashboard, uh there's actually uh so many different things to to launch a a a professional project. So I just I feel like it's really a a very important tool for people to address education and expanding their own uh skill sets in order to be able to be more productive and and more well-rounded, multifaceted. >> That's great. So you know the title of our panel has the phrase don't trust verify in it which is as most people know a fairly core tenant of the Bitcoin ethos. AI models, even the open source ones, but DeepSeek, Gemma, whatever it is, those are still neural nets, which are essentially black boxes, even if they're open source. And that's not to mention Claude and Chat GBT and the rest. How do you apply the So maybe we'll start with you, Charles. How do you apply don't trust verify to a neural network? >> Yeah. Uh, so again, technically they're not even called open source. They're called open weights, those models. Uh, and just like you said, it's they're essentially black box. You don't really know how it gets to that answer based on the prompt that you've uh you've asked. The other thing is, especially when you're looking at the closed models behind APIs, you don't even know, but there's a there's a system prompt. So, there's your prompt, there's the system prompt, and then there's the model. And in the system prompt is where all those big labs, you know, ha have the had the um the censorship that they want. So they say if anyone ask about that topic say you can't answer that or you know essentially even potentially uh keep track of who's asking about that tr topic right so the the the amount of trust obviously is is higher um when you were talking about a closed source model but even as you were mentioning even open source models are not are not perfect on that front. Um so you can't the the answer is you actually can't really trust fully uh what you're going to get out of it. So you have to always kind of keep a degree of of of uh skepticism towards the answer that you're getting. The one thing that you can trust is the rails that you're using to access it. So again, if you're if you're looking at a state stateless u you know ser server like like you guys offer uh and you're paying in Bitcoin and there's no account then at least you know you can't be traced. So on that front that's the part that you can trust and to some degree it's like similar to how you know Bitcoin operates. We don't we don't check every single transaction. We trust the rules. We trust how they operate and you know then we hope for the best from from from that point on and when we have a doubt we can actually go and look but we don't look at every single transactions ourselves. Right. So it's it's the similar logic. >> Yeah. So just to add in I think the I think you're totally right. We cannot actually verify that we cannot verify how the models work how they came to an answer. I think what we can do and what we see getting quite popular on our servers as well is it's a funny name trusted execution environment. Um but essentially what you can at least do there is you can verify the code that's running is what they say they are running. So you can verify that on the provider side who is hosting this model that they don't have any logging in place that they are actually running the model that they say they are running. So I agree it's very hard to have the strong don't trust verify that Bitcoin has with these models. Um but I think the best we can do is trusted execution environments because at least like with chat GPT they could sometimes be replacing the model that they're using with a sort of slightly dumber cheaper version. With trusted execution environment we can verify every step along the way and see like this is exactly the model that's being run. I can verify um you might still not be able to trust the answer like don't don't take the medical advice as gospel um but at least you know what's being run which yeah I think that's the best we can do at this point >> DC >> yeah I would say the you guys are spot on in some ways however the the needs of AI actually are inherently in the opposite direction of Bitcoin's don't trust verify ethos that the larger companies are making you more dependent on using them. They're developing these models to be able to handle more and more use cases so that an end customer is reaching for one particular model to do everything. Uh I think there's an important aspect for people to consider there. If you want to remain sovereign, you can't become too dependent uh on any specific model or any specific uh company that's putting those things out. And so that's a very important thing especially for people starting out uh vibe coding apps. You should think about how much you're depending on any one given model. You should think about how much access that you give to any one given model. And there's a lot of choice there for you. I think that's again another area of education. I think a lot of people who want to build things and are eager to get something done, they they owe it to themselves to educate themselves about how software projects need to be considered. There's a lot of areas you might not even think of. So, I think there's a lot of areas of um contradiction between the trends of AI and that ethos in Bitcoin. And it's up to the individual developer to kind of stand their ground by, you know, making sure they're not developing those hard dependencies and really scoping the AI to specific areas of work uh one section at a time and own your own what you're building. You have to be in charge of it. So let's stay on that topic of sovereignty and you're talking about not putting all your eggs in one basket, one models basket, but there are a lot of Bitcoiners who I mean not just developers of apps but who are just using AI for their day-to-day tasks of course outside of Bitcoin as well. And there's a case to be made that these things are called agents because you're giving up your agency to them and they're they're you outsource your thinking uh to them. So TC, how what is the mental model of Bitcoiner uh should go in with when they're using AI in general for their tasks or whatever it might be, right? Do how do you still maintain your sovereignty without completely outsourcing your agency to these models? You have to remain the decider. You have to actually know what it is that you're building. And that might seem like kind of an obvious question, but what I'm really saying is when you think about something with any level of sophistication to it, there are layers and layers and layers of details that really equate directly to layers of choices. And if you want to remain uh sovereign in any way over what you're building, that means you have to be the one making some of the key decisions along the way. I think the the danger is people really truly vibe coding and trying to oneshot things and allowing the AI not only to strategize uh but make every implementation decision and things that affect everything from your software architecture to the end UI that your users will will interact with. There's so many decisions there and so the number one thing you can do to keep your agency is you got to make sure you're you're the one making all those decisions. So, uh, it's kind of a counterintuitive thing when you feel like you have this powerful tool, you can just give it a simple prompt and voila, magic happens. Get away from the magic. >> So, so question to both Charles and Milan, you're building products that and really your incentive is a lot more people use your products, right? And these are AI enabled products and but there is a risk that the people using your products are giving up their agency right outsourcing their thinking. So how do you deal with that dichotomy? We'll start with you Charles. >> Well I mean I wouldn't really see it as they're giving up their agency. I mean again it ends up being your own decision like you you decide what you want to do and what you want to build. Um, but you know the one way that I look at it is again you want it to be permissionless and private. So if you're using an agent and that agent needs to ask permission from a frontier lab to execute work on your behalf or anyone can ask your agent what did actually you know what did the AY do today? This is where you're losing agency. But if no one can ask that question to your agent or your agent doesn't answer that question and your agent can act on your behalf without asking for permission from anyone, this is what this is the best way you can actually keep your own agency and make sure that your agent is working for you and for no one else. >> Yeah, I'm trying to think because I think right now it's still possible to um essentially verify yourself what the agent is doing and check like does this make sense? But what I'm noticing in code is that in some ways it's getting scarily smart and getting sort of better at programming than some very good programmers, which then makes it very difficult for you to keep that agency, keep being able to to verify what's going on. Um, and sort of in if the models get even smarter, then how are we going to deal with that? So far, the only way that we've really thought of like how can we improve on this is you have many different models. They have different strengths. They can be critical of each other. So even if it's something where it's a oneshot thing that it implemented, you can ask another agent for a first pass review. Um so that if it's something that you really don't understand yourself, at least you have someone else looking at it, which is imperfect, but it does feel like the sort of world we're moving into where these models are just becoming very smart in some ways. Uh so for now, I think you know what TC is saying makes sense. Just make sure that you are the one in control. You are the one that ends up deciding. Uh but I think we're increasingly moving into a world where you might need another layer in between to sort of simplify it down for you. um and tell you sort of what are we deciding on? What what am I actually saying yes or no to here? Um so yeah, >> so I think it was you Charles who who mentioned um one way you can maintain sovereignty is by controlling the financial uh transaction there and I think both you and Milan offer services where you can pay with SATs. Uh so uh do lightning or ecashbased micro transactions can they act as a check on just giving up this agency if you will or can they act as a credential or identity uh when you're working with AI. >> Yeah. So I mean the beauty to me for lightning is actually you don't need um identity. You know again on our platform people come they ask for a service they pay in lightning. I have no idea who you are, where you're, you know, logged in from. Uh, is this your first time on the website? When I see transactions come in, that's all I see. U, so you have to think of it as almost like a vending machine. Like when you go when you go to a vending machine, you press the button what you want, you pay, and you walk away. No one's asking you for your name and to fill out some sort of um, you know, KYC uh, form. Um, so that's that's the the beauty of the protocol around Lightning and offering service this way is that you know for sure that no one's track has the ability to track you. You don't even sign up for anything and there's no um identity linked to your payment. Um, so that's the whole point and in in my approach is it's always been this. It's like there's no account. You don't sign up. you don't there's no way of tracking anything that you do on the on the um the the the website and the service. Um so that's the beauty again. It's pure payment. No no ID. >> Yeah. Know I can only agree with Charles. So on our website you when you land no account is needed. It's just an anonymous identifier. You pay in lightning. You pay in Bitcoin. That's all we see right? Like there's no need to put in a name. Um at this point because we also have more normie style people. It is possible to make accounts. Uh it is also possible to pay with credit card, but it's kind of the option, right? If you want actual privacy, that is not the way you're going to go. Uh you all you need is just a small deposit and you can start using. So yeah, I do think there's a very important um path for Bitcoin and for lightning to make sure everyone can access without giving up that info. So if anyone has used a API access for stuff like chat GBT anthropic in or actually even open router which acts as a front end for a lot of these models there is some hardcore KYC you have to go through you have to take a photograph of your driver's license front and back. Do that selfie thing. Move your head around. TC you're clearly a man who uh appreciates his privacy. Have you subjected yourself through that uh that process? >> I'm happy to say I have not. >> So how do you maintain your privacy while using these AI models which obviously you have to uh pay for somehow right the computer is not free. >> Yeah, I I have used to a great part in the past uh paid paid models. Um, I just haven't done it through the uh the API as much as through uh other methods of authentication with them. >> Welcome to predict. The world is a market. Everything is a market. Every headline moves the line. Every moment is your market. Call the moves. Bet on your instinct, your prediction, your edge. Dual bits predict where everything is a market. U so we're seeing some early signs of AI models I mean this is not rampant yet but AI models transacting or AI agents rather transacting with each other using Bitcoin payment. So is that a future that we're we're going to see where models are doing tasks and communicating with each other and paying each other in Bitcoin and in that scenario how do you manage the reputation right how do you understand the agent that you're transacting with reputation system for agents so Charles what what are your thoughts on that >> um well I mean again to essentially what we're building is exactly this which is like in the future agents will call on SAS for AI for tools of of things that they can't do themselves. So it's the whole um you know pivot that we did uh recently. Um and again the the best way is just for oneoffs. You just want to pay with lightning. You don't want to have to sign up and an agent wouldn't want to sign up. Um so again you don't need ID. You don't need a authentification. The only need you have to kind of trust is is the um uh is the service going to like deliver what I'm paying for? Um but at you know fractions of of of pennies for for transactions the risk is pretty low. Um there's more and more kind of like um registries that are getting created where the the services are listed and potentially people are are are voting. So that's you know that's one answer that I don't fully have yet as to how this was this will unfold in the uh in the future. Um I think maybe Noster is an interesting angle and you're you're probably the best person on the panel to to to talk on that front. uh but that's uh that's how I see it potentially in the future. >> So you are like similar question to you Milan. I mean you see these harnesses open claw and others you have agents that that exist and and they do things. So is Nosta or or some kind of reputation based system uh do do you see the need for that soon? >> Yeah. So, I was thinking about this and I actually don't see as much need for it for a service provider like us because we don't really care about reputation just like we don't care about someone's account or their name or anything. If you pay then we will provide the service. So, I'm not sure in what sense that is what matters. I think what actually matters is having a form a form of money that we can trust. We know it comes in. As soon as it comes in, we know it's ours. And I think by doing it that way, you do away with the need for a lot of the more traditional reputation need that you might have. We don't care about reputation. Uh so I think that might be a much simpler answer. Just trust the money instead of a person's reputation or an agent's reputation. >> TC, do you have any thoughts on that? Agents communicating with each other, transacting with each other, and the need for a reputation system. I mean, I haven't done work with that specifically, but I I really like what both of them said, especially about uh micro payments because it seems to me like that's the only way you're going to be able to actually uh execute those type of maybe high volume APIs without the trust and it's a beautiful solution if that that can all be worked out. >> So, we all know what the fiat AI stack looks like. anyone who's used Chad GBT cla or whatever right it's subscriptions it's your email it's KYC that we talked about what is the Bitcoin AI stack look like Milan >> yeah so I think it starts from a different starting point right there is no need to sign up on a website you just go to a website without an account and you pay in Bitcoin I think that's the most important part I think the other side of it is it's fully permissionless also when you start an agent. So if I have a clawed agent, it's very hard for it to do stuff for me because it's hard for an agent to be involved in the financial system. Whereas if I give my agent access to my Bitcoin wallet with a low limit hopefully like a small Bitcoin wallet, u it can actually transact like if it decides I want to call someone, it will use sets for AI, right? So, uh, I think it's a more native way of doing things by running it on Bitcoin rails than running it on traditional rails. Both in terms of just privacy and ease of use, but also how much power you can give to agents. They're they're bit Bitcoin is just much more internet native than any traditional financial system will ever be. So, agents are internet native. They will want to use Bitcoin to do their actual transactions. I think >> yeah um one thing I'm very bullish on is um local models so very small lock models that you would run on your own computer or potentially even on your on your phone um you know even to this day if you run you can run Gemma 4 on essentially an iPhone and you get performance that are you know that rivals uh top frontier models from a year ago. Um, and so it's one thing that I hope to see in the future and one thing that potentially we're working on is a new product which is essentially an app that would run on your phone that has a local model that makes some decisions and every now and then when it needs more compute, when he needs a more powerful model where when he needs to do things that he can't that small model on your phone itself cannot do, this is when they go and and call for services and pay for it in lightning. Um, and ideally also anonymizes the data before sending it. So if you're asking to write, you know, a cover letter for a job, it would remove your name, your address, it would remove anything that would track you. Send this to the bigger model. The bigger model operates whatever your request was, sends you back that, you know, top-notch cover letter and then your local model reinserts your name and whatever identification and then serves it to you. Um, so I'm I'm very very bullish on bullish on the um the local model and you know they keep coming out and and and they keep getting better and better. So that's that's one way that I see as the the Bitcoin AI stack. >> That's great. So we just have under a minute left. TC the question for you. The year is 2030 maybe. Yeah. Turn of the decade. Chances are a person wakes up and before breakfast they've interacted with their AI a 100 times. What does digital sovereignty look like then? And how is it different from now? >> I think uh what you were just talking about doing local compute on your phone is going to change the game as far as that kind of daily uh dayong interaction with it. I think by the time this technology is mobile with you and moving with you on the go, that's going to really change the possibilities. >> Awesome. Well, folks, we're out of time. Thank you so much for listening. Let's have a round of applause for our panel. My name is Avi Bura. I was the moderator. I am the co-founder of Nos Fabrica. We're building reputation systems for AI agents on Nosta. Thank you all. Every year, this community comes together to celebrate, to debate, to build what comes next. And every year, the stage gets bigger. Sound money, center stage. So, where do you go to celebrate the next chapter in Bitcoin history? You come home. Nashville, July 2027.