Most people mistake a headline for a verdict. They are wrong.
Yesterday, Crypto Briefing announced the launch of an "Internet Court" on Starknet—a decentralized arbitration protocol designed to settle disputes between autonomous AI agents. It sounds revolutionary: a court built for machines, powered by zero-knowledge proofs, on a Layer 2 scaling solution that promises speed and low fees.
But I’ve spent 26 years in this industry, 8 of them auditing smart contracts and building decentralized infrastructure. And I know that a court without a code is just a courtroom without a judge.
Let me stress-test this narrative.
Context: What Is the Internet Court?
The project claims to be a smart contract system deployed on Starknet, leveraging its ZK-Rollup architecture for high throughput and low transaction costs. Its stated purpose? To automate dispute resolution for "agentic commerce": the emerging world where AI agents negotiate, trade, and execute contracts on behalf of humans.
Starknet’s native account abstraction is a logical choice here—agents need to operate as autonomous entities, signing transactions and managing assets without human intervention. The idea is elegant on paper.
But elegance on paper is not a security audit.
Core: The Missing Evidence
As a Senior Security Analyst who audited over 40,000 lines of Solidity during the 2017 ICO boom, I learned one unbreakable rule: trust is not a feature; it is an archived receipt.
This Internet Court has no archived receipt.
- No code. No GitHub repository, no open-source contract to inspect. The entire protocol is a black box.
- No team. No founder name, no LinkedIn profile, no prior track record. The project is effectively anonymous.
- No audit. Not even a mention of a third-party auditor. In my Istanbul Node Audit experience, I refused to sign off on code that lacked a full security review. This project hasn’t even asked for one.
- No testnet activity. No transactions, no deployed contracts on a public test network. Just a press release.
- No users. Zero real-world disputes resolved. No AI agent integration announced with verifiable on-chain data.
The core technical claim—that Starknet’s ZK-Rollup makes this viable—is true in the abstract. But the specific risk vectors are ignored.
Consider the arbitration logic. How does the court verify an AI agent’s identity? Account abstraction allows agents to operate, but it also introduces signature malleability and contract-based authentication risks. In my 2022 bear market liquidity freeze experience, I saw oracle manipulation bring down lending protocols. Here, the oracle feeding evidence to the court—off-chain data from AI models—is a prime attack surface. Without a robust, decentralized oracle and a fraud-proof mechanism, the court is a honeypot.
Trust is not a feature; it is an archived receipt. Until I see a receipt—a verified contract on Starknet, a public audit report, a list of known vulnerabilities addressed—I cannot trust this court.
Contrarian: The Narrative Trap
This is a bull market. Euphoria masks flaws. The Internet Court is being sold as a paradigm shift: "AI agents will need courts, and Starknet is building it." But this narrative is dangerous because it confuses novelty with progress.
Consider Kleros, the existing chain arbitration protocol. Kleros uses a decentralized jury pool, game theory, and economic incentives to resolve disputes. It has been running for years, has audited contracts, and has handled actual cases. Yet it remains niche. Why? Because arbitration is a coordination problem, not a technology problem. The hard part is getting parties to accept the jurisdiction, to stake tokens, to accept the outcome.
The Internet Court adds a layer of complexity: AI agents as litigants. How does an agent prove it was wronged? How does it appeal a decision? Current Kleros-style arbitration requires human jurors to review evidence. If this court replaces humans with automated logic, the risk of bugs or manipulation skyrockets.
History is the only consensus that never forks.
We have seen similar hype cycles before. In 2021, dozens of projects promised "decentralized identity" or "on-chain governance" without shipping anything. Most vanished. The Internet Court has the same hallmarks: a catchy name, a popular narrative (AI + crypto), and zero execution.
Takeaway: Wait for the Receipt
My advice is boring but essential: ignore this announcement until you see verifiable code, a public audit, and at least one real arbitration case executed on-chain.
If Starknet wants to lead in the AI governance space, they need to fund open, auditable tools—not marketing stunts. The true value of blockchain is not speed or hype; it is permanence and verifiability.
An image is fleeting; its hash is the truth.
Let’s wait for the hash.