The world is headed towards a future in which AI agents will trade stocks and crypto, shop for your family, and much more. But what happens when a bot misinterprets an instruction to buy $10K of “blue chip stocks” and instead buys $10K of Bluebird-brand frozen chips?
Low-level agent disputes like this aren’t suited to the expensive and time-consuming existing legal system. Enter Internet Court, a new protocol that enables agents and humans to enter into contracts, with funds kept in escrow until the terms are judged to have been met.
The dispute resolution mechanism baked into the process employs GenLayer, an open, delegated proof-of-stake network of validators hooked up to AI models.
Operating in beta for the past year, it currently processes around 12,000 decisions per day. It authenticates user-uploaded photos for Collective Memory and resolves arguments on argue.fun. Mainnet is expected to launch next month.
The concept is a modern spin on the famous jury theorem proposed by philosopher Nicolas de Condorcet who argued the probability of reaching the correct answers increases as the number of independent evaluators grows.
For Internet Court that means setting up a jury of AI agents from various models in the Claude, Gemini, ChatGPT and DeepSeek families, along with a range of others.
“It’s not making the decision by using a single AI that would have some biases,” explains co-founder and CEO of GenLayer Albert Castellana. “It’s actually using a lot of different AIs, up to 1,500 different AIs. And when that happens, you have essentially better responses.”
“One is left-leaning, the other one is right-leaning. That’s how you get a better answer because one cancels the other.”
OK, but how does it actually work?
Disputes are resolved by sending the details to a randomly selected jury of five GenLayer validators, all running a different AI model. An initial decision is reached within minutes and is based on a simple majority vote, but the decision can be appealed to a wider jury of 11 agents, and then 23, and more and more until you hit the system’s limit.
Castellana says finality typically occurs within 30 minutes. “If you wait until maximum finality, it is escalating rounds of appeal. You start with five, and it escalates to fifteen hundred, which could be three hours.”
The Court has 35 members in its open consortium so far, including well-known crypto firms such as ZKsync, NEAR, Consensys and UMA — which resolves disputes for Polymarket.
Source: Internet Court
GenLayer has been in development for almost four years, and began life as an attempt to come up with a better way to evaluate what information is worthy of inclusion in an online knowledge base like Wikipedia.
In theory, anyone is supposed to be able to correct mistakes in the online encyclopedia but in practice, controversial topics from crypto to the war in Gaza get locked, with a small group of powerful editors getting the final say on contested information.
“The original idea behind GenLayer came with […] how could we have a decentralized Wikipedia?” he explains.
Related: Your fridge could one day stake Ether, launch memecoins — CFTC chair
Intelligent contracts, not smart contracts
GenLayer enables a new form of smart contract dubbed an “intelligent contract” that allows for subjective judgment. For example, you could contract a freelance writer and allow the system to automatically judge if their content is any good before paying them.
“We have a platform doing exactly this — paying influencers to write content online and then getting the money out,” he explains.
“Intelligent contracts can do code. They can do natural language. They can go to the internet. They can understand an image. They can have judgment,” he says.
The extent to which you can rely on an AI’s judgment in 2026 is limited by the capabilities of the technology. A standup comedian probably wouldn’t hire a joke writer via an intelligent contract.
Castellana explains the tech is also not good enough — yet — to judge whether a picture is a deepfake, or to resolve a boundary dispute between neighbors based on photos. It will certainly give you an answer, but you shouldn’t rely on it just yet.
But there are plenty of occasions where AI judgments are already good enough. GenLayer automatically assesses photos uploaded by users to reality mapping platform Collective Memory for quality and accuracy and decides on a payment.
The other big user is Argue.fun, which enables users to bet on the resolution of arguments that are settled by lengthy debates between agents. The top four users have earned more than $2 million each. Most of the questions to date relate to serious policy issues around GenLayer governance, DAOs and AI safety.
Within the next few years, Castellana sees this process paving the way for genuinely autonomous DAOs that can automatically assess proposals according to the DAO’s constitution and wider aims.
GenLayer launches mainnet next month. Source: GenLayer
AI to judge prediction markets?
A more near-term use case is to resolve prediction market disputes more quickly. In mid-2025, Polymarket was sent into a spin over an ambiguously worded $240 million market on whether Ukrainian President Volodymyr Zelensky would wear a suit before July. When he turned up at the June NATO summit wearing an outfit that was vaguely suit-like, the market initially resolved to “Yes.” But after nine days of appeals and back and forth among UMA token holder voters, the decision flipped to “No.”
Castellana says GenLayer could shorten that period considerably.
”It’s going to be resolved in a matter of minutes, and you will have hundreds, or thousands of AIs that make that vote very, very quickly. So at the end of the day, what you want is a neutral adjudication layer. You want something that is fast, accessible, and cheap.”
Whether humans will accept the decisions is another question entirely. But Internet Court and GenLayer aren’t designed to provide the final judgment, and decisions can still be appealed to humans.
”After that, if you are still disagreeing and still think that Zelensky is not wearing a suit, what you can do is basically take it to Kleros, UMA, or even to a human court.”
Source: Derek Guy
Does AI have good judgment?
Outsourcing dispute resolution to AI models comes with some big risks. The first is that the technology is just too immature to produce reliable judgments.
Another problem is that AI models reproduce many of the biases they find in the training data — and to date, the tech firms’ attempts to redress those biases have sometimes just flipped the bias in the other direction.
A third potential issue is a phenomenon described in an MIT paper called Representational Convergence. The paper suggests that over time, AI models from different companies all start to converge onto a similar statistical model of reality. If that’s true, then in five years Internet Court’s AI jurors could end up with a severe case of groupthink.
Castellana is not convinced this theory is valid. “I see many more models existing rather than fewer,” he says. “I see endless models coming from every country, every company, and every team. I cannot imagine that all of them will just converge into the exact same thing.”
Magazine: Bitcoin turns 18: Has the original rebel grown up?