The press release promised a judicial revolution. A dedicated “Internet Court” on Starknet, designed to adjudicate disputes between autonomous AI agents. The narrative is intoxicating: a trustless, scalable, and automated legal system for the coming age of agentic commerce. But when I looked for the assembly, I found only a press release.
The code whispered what the pitch deck screamed: this is a concept, not a protocol. As a crypto security audit partner who has dissected hundreds of launch announcements, I’ve learned to ignore the marketing and read the bytecode. Here, there is no bytecode to read. The project, which I’ll refer to as “Internet Court” for lack of an official name, was announced on March 1st by a crypto media outlet. It positions itself as the first chain-based court explicitly for AI agents, running on Starknet’s zk-rollup. The stated goal is to resolve commercial disputes—think breach of smart contract, failure to deliver computational work, or misuse of funds—without human intervention. The agents are expected to deposit stakes, present on-chain evidence, and accept verdicts enforced by smart contracts.
This is the hook: a beautiful, aesthetic vision of a self-regulating digital economy. But beauty is the most sophisticated rug pull. And every exploit is a story poorly told. So let me dissect the tale.
Context: The Hype Cycle of AI + On-Chain Governance
We are deep in a bull market where the fusion of AI and crypto is the hottest narrative. Every week, a new project claims to bring “decentralized AI” or “autonomous agents” to market. Starknet, with its native account abstraction and low transaction costs, is a natural platform for agent experimentation. The Internet Court taps into this zeitgeist perfectly: it promises to solve the critical problem of trust in machine-to-machine transactions. Without a neutral arbiter, agentic commerce remains a fantasy—who judges when two AI trading agents disagree on a price feed? The press release positions this court as the missing piece.
But here’s what the press release didn’t scream: the identity of the team, the location of the repository, the design of the dispute resolution mechanism, the security assumptions of the oracle or judge selection, the token model, or even a single line of pseudo-code. The only verifiable fact is that it “runs on Starknet.” This is not a launch. This is a tweet storm disguised as a news article.
Core: A Systematic Teardown of the Empty Promise
Let’s approach this with the rigor I apply to every audit. A viable on-chain court needs four components: (1) a dispute initiation mechanism, (2) an evidence submission and verification system, (3) a judge selection and compensation model, and (4) a verdict enforcement engine. The Internet Court, as described, has none of these publicly defined.
Based on my audit experience, the absence of technical details is the single largest red flag. In 2020, I identified a critical integer overflow in a Compound governance upgrade—the team patched it silently. In 2024, I reviewed an AI-agent marketplace and found a prompt injection that allowed agents to steal ten million dollars. Both projects had extensive documentation and code. The Internet Court has zero. That implies one of three things: the technology doesn’t exist yet, the team is hiding intentional flaws, or the project is vaporware designed to capture attention (and possibly funding) before delivery.
Let’s examine the technical assumptions. Starknet’s zk-rollup provides security and scalability, but it does not solve the oracle problem for off-chain evidence. An AI agent might rely on a price feed from a centralized API—if that feed is manipulated, the court must detect it. The press release offers no solution. Furthermore, the notion of an AI judge is fraught with alignment risks. How is the model trained? Who selects the training data? Is there a slashing mechanism for dishonest agents? The silence on these points is deafening.
From an economic perspective, the court likely relies on staked assets or transaction fees. But without a token or fee structure, we cannot assess sustainability. The only plausible model is that agents pay a small fee in STRK to file a case, with the losing party reimbursing the winner. But this creates a system prone to nuisance lawsuits unless there is a significant barrier to entry.
Every exploit is a story poorly told. The Internet Court’s story is missing entire chapters. The risk is not just technical immaturity—it’s the possibility that the entire concept is unworkable without months of R&D and rigorous formal verification. Beauty masks the architecture of greed.
Contrarian: What the Bulls Might Get Right
However, in the spirit of objective critique, I must acknowledge what the proponents of this announcement could be seeing. Starknet’s ecosystem is lean but focused. The foundation has actively funded innovative applications through its grant program. This Internet Court could be one such grant recipient, still in stealth mode, with code that will be open-sourced in weeks. If so, the timing of the media coverage is deliberate—to attract developers and testnet users before the mainnet launch.
Moreover, the underlying need is real. In 2024, I audited an AI-agent marketplace where agents negotiated data access rights. The lack of an arbitration layer forced all transactions to be over-collateralized, killing efficiency. A robust on-chain court could unlock a new wave of trustless automation. Starknet’s account abstraction is uniquely suited for this, as agents can hold their own assets and sign transactions without human intervention.
The contrarian view is that the announcement itself is a signal of intent, not a lie. Teams often announce before they have a working product because they want to build community early. The risk is that the community buys the dream without verifying the foundation. Silence is the only honest consensus mechanism—and right now, the silence speaks volumes.
Takeaway: Demand the Code, Not the Narrative
Truth hides in the assembly, not the press release. The Internet Court on Starknet may one day be a landmark project—or it may disappear into the graveyard of failed DeFi experiments. Until the team reveals its contracts, its dispute logic, and its security audit, the rational response is skepticism. Do not invest in the narrative. Do not trade STRK based on this news. Instead, wait for the repository to go public, read the bytecode, and then decide.
In a bull market, euphoria masks technical flaws. My job is to hold up a mirror to the code. And right now, the mirror reflects only a blank screen. The question remains: will the court deliver a verdict, or is it already guilty of promising what it cannot prove?