AI Agent Disputes Draw OKX, MetaMask to Internet Court

AI agents are supposed to remove humans from transactions. Internet Court exists because removing humans also removes the obvious person to call when a deal breaks.

A 27-firm consortium led by the GenLayer Foundation, with backers including OKX, MetaMask and Matter Labs, has formed Internet Court to handle disputes between AI agents, according to CoinDesk. The pitch is blunt: agent-to-agent commerce is moving faster than the systems that can police it.

Why AI agents need a shared dispute court before they can spend money safely

The assumption behind agentic commerce is that software can negotiate, pay and execute without a person approving every step. The reality is messier. If agents can strike deals, they can also misunderstand terms, contest delivery or act across incompatible systems.

CoinDesk reports that AI agents already negotiate and pay one another without humans in the loop. But the same source says agentic systems lack a way to settle disputes, while traditional courts are not built to handle those cases.

That is the gap Internet Court is trying to fill. It is not being presented as a literal government court. It is a shared protocol layer for AI-based payments, escrow and dispute resolution, designed so different agents and platforms can recognize common rules instead of each building a private process.

“Internet Court is the shared place agents can turn to when a deal goes wrong. Machine-speed money needs machine-speed adjudication,” said David Riudor, CEO and co-founder of the GenLayer Foundation.

The important signal is who showed up. OKX, MetaMask and Matter Labs are not described in the source as having the same role, but their presence points to a clear industry concern: agent-driven finance will not scale if every failed transaction becomes an unresolved argument between machines.

This also connects to the broader AI-agent tooling fight XOOMAR covered in Model Lock-In Cracks as Vercel AI Agents Pick Labs, where agent systems are becoming less tied to one provider and more dependent on interoperable infrastructure.

What GenLayer’s 27-firm group is trying to standardize

Internet Court is best understood as connective tissue. The consortium wants to make payments, escrow, execution, identity and dispute resolution work together for AI agents.

The source names several emerging standards that each solve part of the stack:

The problem, per Albert Castellana, co-founder and CEO of GenLayer Labs, is fragmentation. Each layer can work on its own while still leaving agents to resolve the gaps between payment, identity, contract terms and enforcement.

“Internet Court makes them work together,” Castellana said. “With our founding members, we’re turning a fragmented space into a single open skill that any agent can use to make financial commitments hold up, even when they're contested.”

MetaMask’s role is more specific in the source. GenLayer is using the MetaMask Smart Accounts Kit, including ERC-7710 delegations and its x402 Facilitator, as part of Internet Court, according to Ryan McPeck, Smart Accounts Lead at MetaMask.

For readers tracking crypto payments infrastructure, this sits next to the same practical question raised in Nium Snaps Up Cypher as Crypto Payments Get Serious: crypto payment rails only matter if they can support real transaction flows, not just movement of funds.

How an AI escrow and dispute layer could sit inside a transaction

The source does not publish a full step-by-step technical design for Internet Court. That matters. Any detailed workflow today should be treated as implementation-dependent, not as confirmed product behavior.

Still, the basic design problem is clear from what GenLayer says it is building.

Before Internet Court-style infrastructure, an agent commerce transaction can look like this:

• Before: One protocol handles payment, another handles identity, another handles agent communication, and no shared layer decides what happens if the agreement is contested.
• After: Agents, wallets and apps can point disputed commitments toward a common adjudication layer tied to payment and escrow logic.

The most likely pressure points are easy to see without inventing mechanics. An agent may claim it completed a task. Another agent may reject the result. A payment may depend on whether the output matched the agreed terms. A service agreement may be too small or too fast-moving for a conventional legal process.

CryptoBriefing’s related report, based on the same Friday statement, says Internet Court’s early use cases include AI agent guardrails, automated enforcement of micro-value service agreements and decentralized adjudication of disputed digital evidence. Those examples suggest GenLayer is aiming at disputes that are frequent, technical and too small for traditional resolution channels.

The credibility test will sit in the details: who can trigger a dispute, what evidence is admissible, how decisions are reviewed and how fast funds can be released or withheld. The announcement names the ambition. It does not yet answer every governance question.

Why OKX, MetaMask and Matter Labs want common rules for agent commerce

The business logic is straightforward. If users and developers are expected to let AI agents make financial commitments, the system needs more than payment execution. It needs a way to catch failure.

Matter Labs frames the issue as a complete transaction flow. Vassilis Tziokas, VP of Growth at Matter Labs, said agents will need reliability “from payment all the way through to catching when something goes wrong.” He also said the chain powering Internet Court “runs on the ZK Stack.”

That matters because agent commerce crosses boundaries. A wallet may sign. A protocol may route payment. An identity standard may identify the agent. A separate network may carry communication. If every layer works alone, disputes fall through the cracks.

For developers, common dispute rules reduce guesswork. A team building a shopping agent, trading assistant, booking bot or business automation tool would rather integrate one recognized dispute path than negotiate separate rules with every counterparty.

For infrastructure firms, early standards can become defaults. That is the strategic prize. Internet Court is trying to become the place agents go when a financial commitment is challenged.

A failed AI booking shows the design problem, not confirmed mechanics

Here is a hypothetical, clearly not a reported Internet Court case: a user tells an AI agent to book a refundable hotel under $300 near a conference venue. The agent pays a vendor. The booking later turns out to be non-refundable or far from the venue.

Without a shared dispute layer, the user, the agent provider, the wallet and the vendor may all have different records and incentives. The payment may have moved, but the agreement is still contested.

With an Internet Court-style model, the useful question is not “who feels wronged?” It is whether the transaction terms, agent instruction and delivered result can be compared under a common rule set. The source supports this at a high level because Internet Court is described as combining payments, escrow and dispute resolution, including disputed digital evidence.

Possible outcomes are not specified in the announcement. That is the point. The consortium still has to show how Internet Court turns contested commitments into enforceable decisions that participants will accept.

The trust problem moves from payments to governance

Internet Court answers one obvious weakness in agent commerce, but it opens tougher ones.

Legal enforceability is unresolved in the source. A protocol-level ruling may affect on-chain funds or escrow behavior, but the announcement does not explain how those outcomes interact with courts, regulators or off-chain contracts.

Governance will be just as important. Users and developers will want to know who writes the rules, who reviews disputes, how appeals work and whether large consortium members can shape outcomes.

Privacy creates another pressure point. Dispute resolution can require sensitive information, including prompts, transaction records, business documents or customer data. The more evidence Internet Court needs, the harder it becomes to keep the system safe and acceptable to institutions.

The next practical test is adoption under stress. Internet Court has to be fast enough for agents, cheap enough for small disputes, transparent enough for developers and hard enough to game. If it fails any one of those tests, agent commerce falls back to the same unresolved trust problem that made Internet Court necessary.

Disclaimer: This XOOMAR analysis is for informational and educational purposes only. It is not financial, investment, legal, tax, or professional advice. It does not provide buy, sell, hold, price-target, portfolio, or personalized recommendations. Verify information independently and consult qualified professionals before making decisions.

Impact Analysis

• AI agents are beginning to negotiate and pay without humans, creating new risks when transactions fail.
• A 27-firm consortium signals industry concern that agentic commerce needs shared dispute infrastructure.
• Backers including OKX, MetaMask and Matter Labs suggest crypto and Web3 firms see dispute resolution as key to scaling AI-based payments.