Federico Ast Decentralized justice
How a blockchain-based protocol is prefiguring an Internet-native justice system.
How do you like to introduce yourself?
I am a founder and CEO at Kleros.
How would you outline the trajectory of your life, starting from wherever makes sense to start to where you are now?
I was born in Argentina in 1981. When I was a kid, when I was a teenager, I wanted to be a writer. I loved reading. I loved novels. I read a lot of science fiction. For example, I loved Jules Verne. And then I also loved sports. I was quite good playing football—as an Argentinian, that’s an important skill to have.
And then I went to the university. I did two majors at the University of Buenos Aires, which is a major public university in Argentina—first in economics and then in philosophy. When I was still studying, I started working in online media. Since I wanted to be a writer, I wanted to start my career with was journalism. I started working at this very big newspaper from Argentina called Clarín, which had, of course, the online unit, and in that online unit there was a business section that was for management. It was a bit similar to Fast Company, if you want, but from Argentina. And I started there as a writer and became an editor soon after starting.
It was a one-man show. At that company, it was me doing everything—very Latin American style, very improvised, with few resources. That was the way in which I got in touch with innovation and startups. In those days when I was working there—it was around 2012—people started talking about Bitcoin: magic money from the Internet. Argentina, because of so many problems with inflation and bad governance, was kind of a pioneer country in Bitcoin.
I was interested in the financial aspect, but I started seeing blockchain more generally as a technology that could be used for lots of different things, and in particular to build new institutions. Some people were starting to think about how to use blockchain for building tamper-proof voting systems to prevent voting fraud and stuff like that. And I started thinking, you can use this also for building an institution that covers a very important part of governance of a society or community, which is the justice system.
I was thinking this in the context of Argentina, which was a highly corrupt country. Back then it had lots of high-profile corruption scandals in politics. You would open the journal, the newspaper, and you would always see a big scandal about that. So I was thinking, you have this currency, this technology that you can use for money. And then can you use it to build more transparent governance systems? That was the question I asked myself.
I was also thinking about how the Internet had been helping everyday people from Argentina have more economic opportunities. For example, with YouTube you could watch tutorials and learn how to code or how to do design stuff, and then you could sell your services to other countries and customers in the rest of the world—which was very hard before that, because you could not have sold a $1,000 website to a customer from the US from Argentina.
But there was a problem—it was very hard for you to get paid because there were capital controls, high taxes to receive money from other countries. Bitcoin was a way to circumvent all of those restrictions that the government put on the economic opportunities of people. Now you could learn how to code and contact foreign customers, and now you could actually be paid. But the last thing you need is—if they have a dispute with you because they don’t like your website, and they’re not going to go to court in Argentina against you, or they’re not going to go to arbitration. But what if they could have a very efficient, very fast system to resolve small claims? That was the early idea of how a justice system could be created based on this new technology.
I got tired of working at the newspaper, so I went and I did a PhD. And I did a PhD in business management at IAE Business School, which is quite important in Latin America. And what I researched was moral philosophy and, in particular, how collective intelligence works—how people make decisions together in deliberative settings.
At some point things connected. What if you combine this idea of crowdsourcing and deliberative democracy with this idea of blockchain as a transparency tool? It’s kind of a mini justice system where all of the process is going to be following rules, code, and blockchain, and people are going to be able to participate. Because of course I had studied philosophy, and I loved political philosophy. I had read Aristotle, and I had read ancient history.
I was thinking, okay, you could combine these ideas into building a crowd—when I started I called it the crowd jury, a jury of the crowd. And then I wrote a paper about how I thought that system could work. Pick the jury through blockchain so everything has a rule of law. I thought my contribution would be only to write that paper, that blog post, and maybe someone would read it and maybe implement it. I was thinking, who am I to actually do this? I’m just a poor guy from Argentina—who is going to pay attention to me?
At some point I applied for a scholarship that was given by the government of Buenos Aires to go to Singularity University in Silicon Valley. I won that with this idea of the crowd jury. I went there, and I spent about five months at SU, surrounded by people from different countries, people doing nanotechnology, biotechnology, blockchain, AI, VR—all of the hype technologies.
After that I was finishing my PhD and I had to go back to Argentina. I was thinking, well, okay, what am I going to do with my life now that this Disneyland trip is over and now I have to go back? And I was thinking of finding a job or whatever, and at some point I got contacted by Bitnation. Do you know Susanne?
Yeah, absolutely. Susanne Tartowski Tempelhof. She is hard to forget.
So she sends me an email, and she tells me, “Look Federico”—she knew me because I was in her Slack or something, I mean we had talked a bunch of times because of decentralized governance—and she tells me, “Look, I was at a hackathon and I was a judge, and there was this guy who was presenting a dispute resolution idea similar to what you are thinking. He’s a technical guy, and you are more the business guy. So I think you should meet and see if you can work together.”
Clément Lesaege and I had a Zoom, and then started Kleros—it was May 2017. I only met him in person for the first time, I think, six months later at the Devcon conference in Mexico. That’s how it started. Going to Singularity was important—it was a way for me to be convinced that I could actually work on this and build it, not just write the paper, because I saw people who had started from nothing and built very important projects. Why couldn’t I do this also?
Can you say a bit about what Kleros is and how it has developed?
In the very early days, the main idea was that we were transitioning from this centralized web ecosystem of big companies like Facebook and Twitter—and from a political perspective, they look a bit like monarchies, if you want, absolute monarchies that have centralization of decision-making power. They have users who are subject to the authority of the king. And I was thinking of transitioning to this more republican version of the web with Web3, where users actually have a voice and have a vote. That means they also should be participating as jurors in decision-making and dispute resolution in the platforms where they work.
Back then, we didn’t call them DAOs—I think the idea was more like platform cooperatives. That was the word we used. But I was thinking that these platform cooperatives are like small governments or nation-states—digital nation-states—and we need to start thinking about how to govern them. That what we thought when we started Kleros.
We expected that soon there would be a decentralized Uber, decentralized Airbnb, decentralized e-commerce, and decentralized everything, and all of those would have something in common. They would have disputes between their users, and you would need a system to resolve them. So the initial vision was: what if they plug their DAO into Kleros, send disputes, and then Kleros picks the jurors and sends back the decision to the case? It was a justice-as-a-service project, based on systems of economic incentives. In order to be a juror, you need to have the token and stake it into the court, and then you are drawn randomly through an algorithm.
The system was inspired by how ancient Greek courts used to work. Kleros means “randomness” in Greek. The kleroterion was the machine that they had at the door of the courts in Athens to do the allotment lotteries for jurors. And the token that we use is named after the pinakion, the ancient Greek ID that people used to be drawn. They deposited the pinakion, and it would be drawn by the machine.
You might imagine the reaction when we were pitching this ultra-sophisticated interpretation of the ancient Greek democracy to VCs—but we survived, and so we started implementing all this.
The main problem we have is that the Web3 ecosystem grew way slower than everyone expected. We don’t have decentralized Airbnb, we don’t have decentralized Uber, we don’t have any of that. So we started focusing more on business-to-business and higher-value disputes—prediction markets is one of them, decentralized finance like lending protocols, things like that.
Another big segment is also the enterprise sector—traditional insurance companies, traditional governments, and all that. Even if the enforcement is not on-chain, as we envision this with Web3 customers, Kleros can still produce a decision thay is seen as legitimate by users because it’s made by a jury that’s selected randomly through some transparent method, and they give justification. So this is actually what is working better now for Kleros—this enterprise segment. That it’s not what we imagined in the beginning. In my mind there was always this idea of enterprise and using this in real governments. But since we expected Web3 to grow way faster, we didn’t pay that much attention to it. So we started with it about two years ago, and now we are starting to see the fruits.
That’s it, more or less. Yeah, I can explain. Tell me where you want me to double-click now.
Let’s first double click on that word protocol. On the website right now, if you go to Kleros.io, it says “the justice protocol.” What does the word mean for you? Why is it important, not only as a technical term but as part of the messaging that you use to introduce the project to the world?
I think it’s the set of rules that defines how this system of justice works. We typically call it decentralized justice. So this is a protocol that follows these rules, this procedure, and that also—well, in our case at least—coordinates people in order for them to produce a service that we could call justice.
Some people don’t like me to call it justice, especially those coming from arbitration or the legal field. I still think that it was a good idea, at least in the very early days, to call it justice, because it got people’s attention way more than if we were just building a solution system. And also because, the way I saw this in the early days, it was really a justice system for the metaverse.
There is an interview I did with Max Keiser in 2015, I think, where I said: “Imagine you have—you’re in a virtual world, a VR world. You have a dispute, and my avatar can sue your avatar because of something you did to me in some VR world, and then this can go to a trial, and then you could have a bunch of bots as jurors in that trial.” That’s how I was thinking—I was thinking of this as a justice system. The word protocol came in when people started speaking of protocols in Web3, but also protocol as set of rules that you follow for delivering justice.
How was the protocol developed? Can you talk a little bit about the social dynamics? I mean, you talked about your PhD process initially, but I imagine others got involved. What was the design process like? And how much is that ongoing?
The technical elements for Kleros were developed by Clément, who is my co-founder. I don’t code, but I do have ideas of how I think this should work. For instance, when we started, my early idea was for the mechanism of Kleros to use a matching algorithm. You would have a type of dispute involving, I don’t know, like e-commerce of websites, and then you’ll have an algorithm trying to look for jurors who would be qualified in that type of skills.
When I met Clément, he came with a different idea. He told me, “You cannot do this because you don’t have an identity protocol, so you don’t know who is behind an address. The only way to do this is to use tokens that produce that kind of specialization through economic incentives—that if you stake your token into a court where you don’t have the right skills to resolve the cases, then you will lose money, and this creates an incentive for people not to do that, and then if they do it, they will lose all their money and exit the system.” This is how Kleros started to work in the very early days, and it is still mainly how it works. Since then, however, lots of things happened.
This type of mechanism is okay for some use cases, especially for use cases involving people deeply embedded into Web3, since they are usually okay with this type of financialization. To me, the first time, I found it extremely weird. I didn’t buy it, it was the only way to build Kleros because we didn’t have identity system on-chain. There was no other choice. But then we moved forward, and we launched this project called Proof of Humanity that can solve this problem of having one person, one vote. Now you don’t necessarily need to have the system where people with more money have more votes in the jury, which was a big friction point for many.
And the there was this other idea from Glen Weyl, Vitalik Buterin, and Puja Ohlhaver: you can start having SBTs.
Soul-bound tokens, which attach permanently to a certain address.
Yeah. With soul-bound tokens, version 2.0 of Kleros launched Beta in November, and it can do that. You can pick a panel of jurors who are part of a certain DAO, maybe they have to have skills in JavaScript or websites or whatever. Now you can have a fqr more rich design field, addressing lots more use cases that were not possible to address before.
People see this type of mechanism, where they can select jurors from some community or with some skills, as way more legitimate. And this is also driving more adoption. What happened before is that some people didn’t want to just have this token voting system. The pure token voting is still present. You could even use the pure token voting for AIs. We will have AI courts, and AIs don’t have identity availability. So you can just have them do the decision fully incentivized with a utility function, and then if people appeal, they would go to a higher level court where there are humans selected through soul-bound tokens.
This is how the mechanism is evolving in response to different learnings we have over time. And it is way more sophisticated than what it was at the beginning. It is actually starting to look a bit like what I had in mind before starting Kleros.
Can you say a bit about the story of adoption? When you first started deploying this, what did it take to bring people into participating, say, as jurors or bringing disputes? And what were those early experiences like? How did you draw people in and help people understand how to participate?
The first version we launched was called Doges on Trial—a curated list, a verification system based on Kleros, where people would submit images, and then people would verify them. And if there was a dispute about whether an image was or not a doge, there would be a trial.
That was the first thing we did. The thing that happened there in the early days is that we were mostly at the beginning—this is something very new, something that people don’t trust at the beginning. People think, okay, if you want to use this for a low-value dispute and it gets it wrong, and then it’s $200, it’s fine. No problem, we can try it. But the problem is that this was built on Ethereum mainnet, so it was extremely expensive to use for low-value claims, for consumer disputes.
Because of the transaction costs.
Yes, the cost was very high—dozens of dollars, sometimes hundreds of dollars to solve a $200 claim. So it didn’t make any sense. We had to focus more on DeFi claims.
One of the famous disputes was when there was the US election, the previous one with Biden and Trump, and there was a prediction market. Some people bet for Trump, others for Biden, and you remember what happened on January 6 and all that. The market had to clear, let’s say, on December 29. But on that day there still wasn’t clarity about who had won. Most people thought that Biden had won, but there were presentations in the courts up to change the result. So some people claimed that the result for the prediction market was invalid because there was no clarity. So there was this case about who was right. Did Biden win, or is this invalid? People were doing constitutional interpretations, hiring lawyers. There was $2.5 million at stake in total.
It was kind of a strange thing, because these random people from the Internet, through this token mechanism, were deciding something that not even the most qualified lawyers in constitutional law could. It was a really great experiment. But we didn’t see that as how Kleros is going to help the world be a better place. This is something that solves a need for some specific type of use case—prediction markets—which is fine, but it kind of didn’t resonate with the reason why I got into this in the first place.
The reason I started this was the lack of access to justice for the guy who wants to sell his services to a customer, or the guy who gets screwed by a company and isn’t going to go to small claims court because it’s too expensive, and maybe that doesn’t even exist in countries in emerging economies. So all of those high-value cases, even if they are legitimate use cases, were not the most interesting cases to me. But a lot of people came from that because there was money to be made. Because of the transaction fees, you could resolve those high-value cases but still not the small-value ones.
Kleros clearly has a very big community of people who are from the fields of law and arbitration. Many of them haven’t even used Kleros, but they all know about Kleros because it’s kind of a cool thing to know if you are into legal innovation, and you have to have an opinion about the use of economic incentives in arbitration. You can see, if you search for Kleros in Google Scholar, there are lots of mentions, mostly from the legal side, legal tech people. They don’t bring lots of users, but they do bring lots of buzz. If you go to a legal innovation conference, someone is going to be speaking about Kleros.
How much do you know about the jurors, about who these people are?
In theory, we don’t track anything. So we don’t know who are, and in the long run we will know even less. But in the short term when we have—say, we’re working with a fintech company from Argentina, and they send disputes to be resolved, and to resolve it you need to understand Spanish, and you need to understand how crypto works, stuff like that. So when we start a new use case, we tell people from the community, “Okay, who knows how to resolve this type of case?” Then they can stake into this court, and they will be drawn.
So even if we don’t know who is behind every address, we generally know from which part of the community they come. Sometimes some of them tell us, “I was this juror, this is my address,” and it’s okay. We never ask, and we don’t have any tracking. Some of the community are crypto hardcore, who value privacy above everything. Some of them are lawyers, and they say, “Well, it’s not a problem, this is my address.” So we have a bit of both.
Have you had experiences of capture—experiences, for instance, where you saw the protocol used in a way that you didn’t design it for and didn’t expect, or where a source of power outside of the protocol was able to use it for its own ends?
Fortunately not yet. There was an attempt in Case 1170. There was this attacker called Avraham Eisenberg, who is in jail because of an attack on a protocol called Mango Markets, and he tried to do something like that to Kleros, creating a narrative about Kleros being rigged. I think he hired some influencers to try to attack the system in a way that is a bit similar to what you would expect when there is a very high-profile trial in the real world. The OJ Simpson trial—everyone has an opinion about this, and if the jury rules in a different way to what was expected, people will think it was rigged, even if it was not. So there was an attempt of an attack similar to that. It didn’t succeed.
At some point I think there’s going to be things like this. We don’t want—of course we didn’t build this for people to use it in assassination markets or stuff like that. There is a supreme court in Kleros: the General Court has some rulings about morality, where if you have a case about assassination markets or something, jurors are instructed to refuse to arbitrate. But in the end, we don’t control it. Someone could come, copy the code and build another version of Kleros, and do whatever. I don’t encourage that—I strongly discourage that. But that is the world of open source.
Has it happened where someone has tried to fork it?
No, because it’s still a bit early-stage. I think we still need to prove that this works as we expect it to work. But I think at some point, if we are successful—that is the irony of success. If we are successful in building this, I’m sure someone will try to do that.
Are there cases or stories where you feel like you’ve had that glimpse of success, where you’ve seen it working the way you really wanted it to?
Yeah, we have a bunch of those. One of those was in 2020. There was this app that we built called Kleros Tokens designed to analyze and vet cryptocurrencies that people might want to list. It was after the initial coin offering boom, and there were lots of scams. We built an app where if you want to list your token on an exchange, you can submit it and a jury will verify it. If you make it through the jury you get listed. It’s a good tool to prevent scam projects.
There was one case about a coin called Baer Chain. People sent the submission and it was challenged, and it went to trial. When the jury started to analyze the case, they saw that the whitepaper of this currency had a CTO called Scott Bingley, who claimed to be affiliated to some Oxford blockchain group. They called Oxford, asked for this guy, and someone answered, “No, we have no idea who this guy is, and there is no Oxford blockchain-whatever here.” It was a fake profile with a stock photo—it was a scam.
Six months later, we read on CoinDesk that the police from China made an arrest in a project called Baer Chain because of a Ponzi scheme or whatever. So Kleros got that information six months before the Chinese police. Imagine all of the people that got scammed during those six months—if there was a robust tool for the certification of tokens, this could have been prevented. That was an amazing thing.
The second moment is happening now. We are working with a fintech company, and we developed this system where you have a user that has a claim against the company—it’s like Coinbase, it’s the Coinbase of South America. They have a claim on the company because they were charged for something they didn’t buy. They go to customer service, and then customer service says no, and they can just go to small claims court. But people don’t go because it’s very cumbersome.
What we did with this company—developed a system where, after customer service rejects a claim, they offer going to Kleros, and the company will pledge to abide by the decision. But if the customer loses, they can still go to small claims court, so they don’t lose any access to justice. It’s asymmetric.
We have had already a hundred cases like this, and 90% of people who go to the system and lose their case still are happy with this process and stay as customers of the company. This is quite amazing because this is what I envisioned. This is what people feel when they are treated fairly by the company. They got their day in court. It only took four days and not four years or four months. And then they got people from the community—peers who analyzed the case and thought that they were not right. And they go, “I’m seeing five of my peers saying I’m not right. Okay, maybe I’m not right.” And then they come back to the company, stay as a customer.
So for the company, the person stays as a customer—customer retention. And more importantly, they don’t go to the competitor, and even worse, go to Twitter to say, “This company, they are scammers, don’t use it.” So there is a reputational risk. So I’m seeing this working perfectly well now, in the way I intended it to work, in the sense that it creates value for everyone. There is value for the company because it avoids losing reputation. There is value for this person who has access to justice in a few days, at a cost that is nothing. This is what I think should happen, this is how it should work, and that’s what we are trying to scale to other industries now.
What would it take to have real kind of accountability in the blockchain ecosystem? Because one thing that was really striking to me in the 2021-22 scam wave, and the collapse wave of things like FTX and Terra-Luna and so forth, was how immediately things went to territorial courts. It was striking that the crypto ecosystem was really not equipped to handle the scale of bad behavior that was going on, and people were immediately running to the legal system that supposedly this whole ecosystem was built to get out of.
That’s a very good question. I think that Kleros and all the projects for dispute resolution are going to be a self-regulation tool for the ecosystem. Like the tool we did for verifying tokens—if we as an ecosystem can verify tokens and root out the fake tokens, then we will have fewer scams and less need for outside regulators to intervene.
For now, it is too early to use Kleros as a binding arbitration system for all kinds of disputes. The FTX thing was too complex, too many moving parts. You could resolve that through arbitration eventually, and maybe not even that one, because it was so complex. But there are other cases. Binance had cases that were sent to arbitration—though it was a traditional arbitration process, not a crypto-native method.
Arbitration is meant to be voluntary by parties. So lots of protocols have arbitration clauses. And sometimes you don’t know what goes through arbitration because it’s confidential. I think for now we are still a bit early. But if we succeed in this—I’m not saying that we’re going to replace all of the traditional arbitration. I think for more complex cases, I people will still want to go to courts or traditional arbitration. But what I care more about is disputes happening for a few thousand dollars—or fifty, or a hundred—where people will not bother. If you want to go to traditional arbitration, it’s at least $1 million, otherwise it’s not worth it. But all of the small economic activity happening in blockchains, like the contractors making videos or websites, that is for a few thousands, and sometimes parties are anonymous. Sometimes they are in different jurisdictions. All of that, I think, Kleros can handle. I don’t know about the high, multi-million dollar cases—I think we are still a bit far from that.
You talked a bit about the influence of ancient Greek democracy on the design of this. Are there other legacies that you’ve drawn on, other kinds of historical patterns or lessons that you’ve drawn on in designing this protocol?
The obvious influence is the Greek. But then there is another part of Kleros that is—this is even more science fiction—a supreme court for the Web3 ecosystem. You have disputes happening or governance proposals happening in a protocol, and then someone claims that this breaks the fundamental values of the DAO, and they want it to be overturned. This is the type of things that the supreme court does.
The way we structure the Supreme Court of Kleros—the mechanism is inspired by the Greeks also. It’s a system called the nomothetai, the law-makers. But another big inspiration for me is the Founding Fathers of the United States. All of the discussions they had, the separation of powers—I read the Federalist Papers, I read Madison, all of that. And also I read the people that they read. I always see ourselves as trying to build institutions for the Internet age, just as they saw themselves to be building institutions for the industrial age. They only had the Greeks and the Romans to learn from, and we have a bit of that—we also have what they did.
We are also pioneers trying to understand new things going on around us. The challenge we have is needing institutions that can be resistant to the age of AI, because the vectors of attack are going to be way higher. Everything happens way faster, so we need as much help as we can get.
I also read a lot about random selection in general, and about the medieval Italian city-states like Venice—how they chose the Doge of Venice, with all of these super complex mechanisms. And also the Florentine republic, and why it failed according to Machiavelli’s understanding and others.
The big question that we have—which they discussed in that time—is whether a good community is made of good laws or of good people? I also read a lot of social capital stuff—Robert Putnam’s Bowling Alone, and the study he did about the Italian regions, where the same institutions have different results because of the underlying social capital. So I read a lot of history to try to learn about patterns that we can use.
I also read a lot of things about game theory to try to understand how eventually AIs will build their motivations. Part of that is behavioral game theory, which is about how people react to situations. For example, the ultimatum game—you have to split some value with others. If you think that the other guy is purely rational, and you have to split $100, you can offer one dollar. He will take it because it’s more than zero—an AI should behave like that. But humans rarely take less than $30 because they see it as unfair.
Finally, what lessons might you share with people who are also building protocols, who are coming after you?
I think that you need to be someone who has a wide view on things, because this is not just computer science. Kleros doesn’t only need to work, people need to see this as a legitimate thing that they want to use. You cannot understand this from just coding—you need to learn lots of anthropology, sociology. My own background has one part that is economics, then the other part is philosophy, and lots of cultural studies. In the end it’s about understanding that all of these protocols are trying to affect human behavior and coordination in different ways.
I always say this—they need to read more books. They need to read the classics. They need to read political science. They need to understand why the Roman Republic became the Roman Empire, and why Athens failed in the war against Sparta, and what were the concerns for the Founding Fathers. You have to learn a lot of history, a lot of politics and economics. I also read the history of economic thought. But these things are not on the reading list of most builders of protocols, unfortunately.