The Internet Court Standard: Three Institutions, Zero Code

Three institutions announced a standard for AI agent dispute resolution. No code. No testnet. No economic model. Just a press release. That is the anomaly. GenLayer, OKX, and MetaMask jointly proposed ‘Internet Court’ — a framework to settle disputes between autonomous AI agents in digital commerce. The narrative is seductive: AI agents trade, lend, and negotiate autonomously. When they disagree, a human-mediated arbitration layer is too slow. Internet Court promises instant, algorithmic justice on-chain. This is not a new problem. Kleros and Aragon Court already provide on-chain arbitration using human jurors and crypto-economic incentives. Kleros has processed thousands of disputes with a TVL around $50M. Their model is battle-tested. Internet Court replaces humans with AI models. That is a fundamental shift in security assumptions. Let me be precise. An AI model — likely a large language model or a specialized decision network — will act as the judge. The model’s decision is recorded on GenLayer’s blockchain and executed by a smart contract. Sounds elegant. But forensic skepticism demands we examine the failure modes. Based on my audit experience, AI models are black boxes. They exhibit bias, hallucinate, and can be adversarially manipulated. If a rogue agent feeds poisoned data into the training pipeline, the model’s rulings become unreliable. There is no cryptographic guarantee of fairness. The ledger bleeds where code is silent. The announcement provides zero technical details. No whitepaper, no repository, no open-source code. The only evidence is a partnership logo list. This is not a protocol. It is a marketing campaign. Let me quantify the risk. The technology stack depends on three unverified components: GenLayer’s underlying blockchain (assumed to be secure but unproven), the AI model’s integrity (unknown), and the incentive mechanism for honest AI operators (undefined). Each component has a high probability of failure. Multiply them, and the probability of a working, secure system is below 30%. Compare to existing solutions. Kleros uses random selection of jurors and a multi-round appeal process. It has survived multiple stress tests. Internet Court’s AI model is a single point of failure. No appeal mechanism is described. No slashing conditions are mentioned. Chaos is just unquantified variance, but here the variance is unmanaged. Now the contrarian angle. Retail traders will see this as bullish for GenLayer and for the AI narrative. They will assume that OKX and MetaMask’s involvement signals imminent adoption. Smart money sees the opposite. These are distribution partnerships, not technical collaborations. A wallet can display a dispute resolution option without any guarantee that the underlying logic is sound. Skepticism is the only viable alpha. The real winners from AI dispute resolution will not be new standards. They will be existing platforms that integrate AI as an optional feature while retaining human oversight. Kleros could add an AI pre-filter for simple disputes, backed by human final appeal. That is a defensive move with low risk. Internet Court is an all-or-nothing bet on unproven technology. I have seen this pattern before. In 2017, I manually audited 50 ICO whitepapers. 12 had fundamentally flawed tokenomics or copy-pasted code. The pattern: bold claims, missing details, strong brand names for credibility. Internet Court fits the template. The absence of a technical blueprint is not a sign of confidentiality. It is a sign of immaturity. What is the concrete path forward? GenLayer must publish a detailed technical specification. The AI model must be open-source and auditable. An economic model for staking and slashing must be designed and stress-tested. A testnet with live disputes is essential. Without these, the standard is a placeholder. Takeaway: Ignore the noise. Set a calendar reminder six months from now. If no testnet with verifiable code exists, treat this announcement as a marketing campaign. Survival is the ultimate performance metric. Until then, let others chase headlines. I will wait for the data. Three institutions announced a standard with zero code. That is the only fact worth trading on.