GenLayer is preparing to launch its Internet Court dispute-resolution system on mainnet in Q4 2026. The AI-powered platform assigns contract disputes between agents and humans to juries made up of different language models.

How GenLayer’s AI jury works

When a dispute enters the Internet Court, its details are sent to a randomly selected jury of five GenLayer validators. Each validator operates a different AI model, an approach intended to prevent the quirks or biases of any single model from determining the result.

If a ruling is challenged, the panel can expand. According to the research findings, the jury can grow from five validators to as many as 1,500 during an escalation process.

GenLayer calls the consensus mechanism “Optimistic Democracy.” It treats the initial decision as valid unless someone contests it, at which point a substantially larger group is brought in to review the result. The optimistic pattern is familiar elsewhere in crypto — optimistic rollups, for instance, likewise treat results as valid until challenged — and applying it here means only contested rulings trigger the step of convening a much larger panel.

The system is designed to support what GenLayer calls “Intelligent Contracts.” Unlike conventional smart contracts, which execute deterministic instructions in the same way each time, Intelligent Contracts can process natural language, retrieve data from the web, and make subjective judgments. That functionality is intended to address disputes such as whether a delivery was “satisfactory,” a question that cannot easily be resolved by standard code alone.

Intelligent Contracts run in GenVM, a custom execution environment built to support large language model functions.

Testnet activity

The Internet Court is currently operating in beta on GenLayer’s Bradbury testnet, named after the science fiction author Ray Bradbury. Research findings place its throughput at between 12,000 and more than 25,000 adjudicated decisions per day.

The testnet is also processing roughly 300,000 transactions daily, a figure that puts adjudicated disputes at only a small fraction of overall testnet activity. About 70 professional validators are active, while the architecture is designed to support as many as 1,500 validators.

GenLayer is also developing the GEN token. The token is planned to have a genesis supply of approximately 4.2 billion and is intended for staking, gas fees, and governance. Because validators serve as the system’s jurors, staking the token would be tied to participation in the jury process.

From YeagerAI to a broader consortium

GenLayer began as YeagerAI in 2024. The project’s testnets have been named after well-known science fiction authors.

According to the research findings, GenLayer has assembled a consortium of between 27 and 35 entities. Named partners include OKX, ConsenSys/MetaMask, and ZKsync.

The consortium’s stated objective is to establish an open Internet Court standard for agentic commerce. The proposed scope extends beyond contract formation to include payments and identity management.

The project’s schedule has changed previously. GenLayer had targeted October 2026 for the mainnet launch, but the launch had not occurred as of November 2, 2026. It is now expected in Q4 2026. Against that backdrop, the mainnet launch itself, the GEN token’s planned staking, gas, and governance roles, and the validator network’s expansion toward its 1,500-validator design are the roadmap elements tied to the revised window.

A proposed layer for agent-based commerce

GenLayer’s design is based on a future in which AI agents negotiate, sign, and settle agreements with one another and with humans, often without direct supervision.

The consortium provides an indication of interest in a shared adjudication layer rather than separate platforms developing their own dispute systems. Its participants include OKX, an exchange; ConsenSys, the company behind MetaMask; and ZKsync, a layer-2 network.

Security is another central consideration. A system in which verdicts can determine the movement of money could attract attempts to manipulate inputs or concentrate validator stake. GenLayer’s architecture is designed to expand from an initial group of 70 professional validators to as many as 1,500 validators operating different AI models.

Source: CryptoBriefing.