Polyhedra launches EXPchain for artificial intelligence applications, analyzing the necessity of AI model on-chain and decentralized zk proof generator

Blockchain interoperability infrastructure Polyhedra, after experiencing a drop in token price and failing to secure the $ZK ticker against zkSync, has recently regained its momentum and launched the "Chain for AI" project, known as EXPchain. The project proposes the Proof of Intelligence (PoI) concept, creating a tamper-proof and trustworthy blockchain for Artificial Intelligence (AI) models. Whether the combination of zk and AI transformation will be successful remains to be seen.

Traditional AI regulation involves sensitive data, zkML as a new solution

The official defines EXPchain as a blockchain protocol designed for scalable, verifiable, and privacy-focused AI applications. As the "Chain for AI", EXPchain integrates Zero-Knowledge Machine Learning (zkML) and a novel Proof of Intelligence (PoI) framework. Key innovations include the efficient zk proof system Expander, and the developer-friendly zkPyTorch toolkit that integrates zkML into traditional AI workflows.

AI plays an increasingly critical role across industries, from using facial recognition to unlock phones to AI-driven loan applications and medical diagnoses. While these technologies offer great potential, they also present challenges. How to ensure AI systems operate fairly, accurately, and securely? How to protect sensitive data without compromising transparency and accountability?

Governments are also working to regulate AI, such as the EU's AI Act and the NIST's AI Risk Management Framework. The problem with traditional approaches is the need to disclose proprietary models or sensitive data, leading to trade-offs between security, privacy, and trust.

Zero-Knowledge Machine Learning (zkML) offers a different solution, where the properties of zero-knowledge proofs can mathematically verify AI systems while protecting data and model privacy. Polyhedra's EXPchain, built on zkML technology, not only addresses AI behavior and compliance, but also provides a scalable and secure verification mechanism.

Technical debt continues to grow, chaining AI transactions benefits accountability

A study shows that the technical debt (the compromises made during software development to quickly deliver or meet short-term needs, which often increase long-term maintenance costs) in the US reached $2.41 trillion in 2022. Additionally, a study by PricewaterhouseCoopers (PwC) indicates that AI is expected to contribute up to $15.7 trillion to the global economy by 2030.

As AI scales, it may exacerbate the expansion of technical debt. The business column Raconteur questioned whether companies are prepared to bear the cost of AI failures, which can include incorrect outputs, data breaches, and cyber-attacks. Beyond economic losses, these errors often harm individuals.

For example, inaccurate data output may lead to machine judgment errors or biased decision-making. Therefore, it is necessary to ensure that every element of the AI-driven transaction, from data input to model output, is verifiable and accountable. Addressing these risks is crucial as we unleash the full potential of artificial intelligence. This is where EXPchain, the AI real-time verification blockchain, comes into play.

Three Major Technical Innovations: Can Polyhedra Solve the zk Proof Generator Dilemma?

Technical innovations include Expander, ExPos, and zkPyTorch.

Polyhedra: Expander is the Fastest zk Prover in the World

The data provided by Polyhedra includes:

• Processing VGG-16 images on a single-threaded CPU in just 2.2 seconds
• Processing Llama-3.1 8B, one token per 150 seconds on a single-threaded CPU
• Performance improvements of four orders of magnitude over previous data

These advancements significantly reduce the cost and latency of AI verification, supporting a variety of applications from private inference to model auditing. Expander also aligns with Vitalik Buterin's vision for the ultimate zk solution.

Layer 2 is mainly divided into Optimistic Rollup and zk Rollup, and for most zk Rollup public chains, ZKP proof generation is a bottleneck, and companies must deploy powerful machines with TB of memory to handle the large number of transactions in ZKP. In the paper co-authored by Tiancheng Xie, the former technical lead of Polyhedra, and Jiaheng Zhang, the chief scientist, they explore a new solution using fully decentralized ZKP to improve the scalability of zk technology.

ExPoS: Extended Proof of Stake

ExPoS is a Proof of Stake mechanism developed for the zkML technology in EXPchain, which unifies and connects all the staking mechanisms on the blockchain into a cohesive staking network, while verifying the behavior and compliance of AI applications without revealing proprietary model data.

zkPyTorch: Developer-Friendly Toolbox

zkPyTorch automatically converts PyTorch operations into zk circuits, bridging the gap between traditional AI development workflows and zero-knowledge machine learning (zkML). This integration allows developers to use familiar tools while significantly reducing the time and complexity of deploying AI applications with zk support.

zkML Can Verify LLMs Under Privacy Constraints

The core of EXPchain is zero-knowledge machine learning (zkML), which supports encrypted verification of AI models, ensuring security and accuracy throughout the machine learning lifecycle, including:

• Verifiable inference: Proving AI outputs without exposing the model or data.
• Model auditing: Verifying the fairness and compliance of performance based on test sets.
• Training verification: Ensuring compliance without revealing sensitive inputs.

Specific zkML applications include:

• Adding digital watermarks to large language models (LLMs). Digital watermarks are subtle and imperceptible features embedded in the text generated by LLMs, used to identify whether the text was generated by a specific model, preventing content forgery and abuse.
• Ensuring model compliance, such as compliance verification in financial institutions.
• Enabling secure multi-party computation in privacy-sensitive industries.

EXPchain's zkML digital watermarking is currently capable of verifying large language models such as Llama-3.1 8B.

Polyhedra's Chief Cryptographer Has an Impressive Background, Driving the PoI AI Proof Chain

EXPchain can be seen as a Proof of Intelligence (PoI) system, creating a tamper-proof and trustworthy blockchain to verify the origin, authenticity, and ethical compliance of AI models. This framework protects intellectual property and ensures transparent accountability by cryptographically linking the source and performance of each AI model to verifiable on-chain records, providing unprecedented transparency for the AI-driven ecosystem.

The driving force behind all this is Zhenfei Zhang, the Chief Cryptographer of Polyhedra. He has previously worked at industry leaders such as Algorand, Espresso, Ethereum Foundation, and Scroll, and is highly respected in the cryptography field. His paper "ZEN: An Optimizing Compiler for Verifiable Zero-Knowledge Neural Network Inference" discusses verifiable machine learning.