In cryptography, we have a sacred rule: never roll your own crypto. Yet last week, China and 29 nations rolled their own AI governance body—the World AI Cooperation Organization (WAICO)—and deliberately excluded the one toolkit that ensures verifiable transparency: blockchain and zero-knowledge proofs. I spent the past three days dissecting the WAICO announcement, not for its political theater, but for its cryptographic blind spots. The result? A governance structure that claims to be "responsible AI" but refuses to let the math speak. It is a decision rooted in opacity, not security.
The Math Whisper That WAICO Ignored
WAICO is a government-led initiative to set standards for AI development, ethical guidelines, and deployment. Its founding members span from Beijing to Cairo, a coalition of emerging economies seeking a counterweight to Western AI dominance. The official document is clear: "Blockchain and cryptocurrency technologies are explicitly excluded from the scope of WAICO's governance framework." No rationale is provided.
To a technologist, this is like building a bank vault but forbidding padlocks. Blockchain, at its core, is an immutable audit trail. Zero-knowledge proofs (ZKPs) allow one party to prove a statement is true without revealing the underlying data. Applied to AI, they can verify that a model was trained on a specific dataset, that its outputs meet certain fairness constraints, or that it hasn't been tampered with—all without exposing the proprietary weights or user inputs.
I've seen this work firsthand. In my 2020 audit of Uniswap V2's liquidity pools, I learned that trust is not given; it is computed and verified. The same principle applies to AI. Without cryptographic verification, you're left with promises and paper audits—the same model that failed during the Terra crash, where code was the only witness to the death spiral. WAICO's exclusion of crypto is not a technical oversight; it's a design choice favoring control over transparency.
Core Insight: The ZK Gap in AI Governance
Let's trace the technical failure. An AI governance body needs to ensure compliance across thousands of models, from recommendation algorithms to facial recognition systems. Traditional methods involve third-party auditors who review logs, inspect training data, and test outputs. But these auditors are fallible, expensive, and—in the context of state-run bodies—often politically aligned.
Zero-knowledge proofs offer a superior alternative. Consider zk-SNARKs for verifiable inference. A model owner can generate a proof that for a given input, the output matches the model's logic without revealing the model itself. This is not sci-fi; projects like ezKL and ZK-ML have been running on testnets for months. In 2024, at the ZK Educational Summit I organized in Taipei, we demonstrated how a hospital could prove that its diagnosis AI meets regulatory accuracy standards without exposing patient records. The audience—500 researchers and developers—watched as the circuit compiled; the proof took seconds.
WAICO's exclusion means that models deployed in its member states will rely on centralized audits conducted by politically appointed bodies. This creates three structural risks:
- Model Tampering Without Detection: Without cryptographic hashes chained to a public ledger, a government could quietly update a model's behavior—say, tightening censorship filters—without any external record. Blockchain's immutability is the only witness.
- Data Provenance Blindness: ZKPs can prove that training data was sourced ethically (e.g., no copyrighted images used). Without them, WAICO must trust the word of developers who have every incentive to cut corners. The math whispers what the network shouts, but WAICO is covering its ears.
- Audit Manipulation: Centralized auditors can be pressured, bribed, or replaced. A public, verified proof cannot.
This is not a hypothetical. In 2022, after the Terra collapse, I spent weeks reverse-engineering algorithmic stablecoins to show how the lack of transparent on-chain verification allowed the death spiral to accelerate unnoticed. If WAICO's AI models suffer a similar failure—say, an algorithm mistakenly denies loans or flags innocuous content—there will be no cryptographic trail to analyze. Only silence.
Contrarian Angle: The Exclusion as a Signal
Here's the counter-intuitive take: WAICO's exclusion of crypto might be the strongest signal that blockchain's transparency threatens the very power dynamics the organization was created to protect. These 30 nations are not technology-agnostic; they are seeking to codify AI governance that centralizes control. Cryptographic verification decentralizes trust, making state-level oversight optional. From their perspective, that is a feature to be feared, not a bug to be fixed.
Perhaps this separation is, in the long run, a gift to the crypto ecosystem. By officially declaring itself incompatible with state-led AI governance, the crypto industry is forced to build its own parallel infrastructure—one that is permissionless, trustless, and global. I've seen this pattern before: when regulators banned certain DeFi protocols, it only accelerated innovation in decentralized exchanges and privacy-preserving finance. The same may now happen with AI verification. Projects like Bittensor, Render Network, and Akash will find a captive market among developers who want transparent governance without asking for permission.
The irony is thick. WAICO's silence on crypto speaks louder than any press release. It reveals that they understand the technology's power—enough to try to contain it. But in a decentralized world, we don't need their permission to verify. We just build the proof.
Takeaway: The Unavoidable Math
The math whispers what the network shouts. WAICO chose to ignore the whisper, but the network does not care. As a researcher who has spent years auditing smart contracts and teaching zero-knowledge proofs, I know that trust is not given; it is computed and verified. The true governance of AI will not be decided in a conference room in Beijing—it will be written in circuits and verified by millions of nodes. That is a future WAICO cannot exclude, no matter how loud they shout.
Proving truth without revealing the secret itself is not just a cryptographic trick; it is the only path toward AI systems that we can actually trust. And that path is already being built, one circuit at a time, outside any government's walled garden.