Market MakingVenture CapitalAbout Us

Revolution in data transmission

Jan 11, 2024
/Table of contents

Marketing Batman

AI and blockchain technologies may seem contradictory, with blockchain focusing on cryptography and decentralization, and AI on statistics and machine learning. However, there’s a growing convergence between them. One notable example is the application of Zero Knowledge Proof (ZKP) cryptography, particularly in machine learning (zkML). The advancements in ZKP, driven by blockchain use, make zkML an intriguing field where cryptography enhances AI.

zkML, or zero-knowledge proof in machine learning, resolves two primary concerns.

  1. It addresses the need for privacy in handling ML model inputs and parameters, safeguarding sensitive information such as personal, financial, or medical data.
  2. It ensures the ability of downstream systems, like on-chain smart contracts, to verify the accurate processing of inputs leading to public outputs.

The key mechanism of zkML lies in training models without exposing underlying inputs and parameters. This approach allows users to maintain the confidentiality of their information while ensuring that the computer executes the correct operations.

Interaction between ZKML user and traine

As depicted in the illustration, users utilizing ZKML models have assurance that their data is secure and that the outcomes produced by the models are valid.

ZKML Oriented Projects in the Market

Modulus ensures the reliability of AI data for web3 protocols through a cryptographic verification system, preserving decentralization and privacy using ZKML techniques. As a specialized ZK prover for AI, Modulus serves as a crucial tool in the era of increasing reliance on artificial intelligence, addressing the potential manipulation of AI feeds by providing robust verification.

Similar to a Twitter verification badge, Modulus aids blockchain users and web3 developers in distinguishing accurate AI-generated feeds from potentially inaccurate ones.

Spectral introduces a novel application scenario involving the utilization of inference feeds for smart contracts. Unlike conventional oracle networks primarily designed for data feeds, particularly for price data, Spectral addresses a notable void by facilitating an unparalleled stream of high-quality inferences.

Through the incorporation of ZKML, Spectral enables modelers to validate the accuracy of their models mathematically, all while safeguarding the confidentiality of the proprietary methodology behind the inference. This certification process ensures the integrity and performance of the models while simultaneously upholding the intellectual property rights of the modelers.

Ion Protocol employs a trustless risk engine powered by zero-knowledge machine learning (ZKML) to assess slashing risk based on the security asset type. This occurs within a zero-knowledge circuit, ensuring off-chain data analysis with on-chain verifiability. By leveraging ZKML, Ion executes complex computations off-chain, avoiding prohibitive on-chain gas costs.

The protocol’s ZKML frameworks enable verifiable inferences, confirming the accuracy of risk ratings. The risk engine determines maximum loan-to-value (LTV) and interest rate parameters for the collateral type, allowing Ion to dynamically adjust interest rates. This trustless mechanism enhances protocol health under various market conditions, optimizing lending outcomes for users.

The Rockefeller Bot, nicknamed Rocky, stands as the world’s pioneering “fully on-chain” AI trading bot, embodying trustless and autonomous attributes akin to decentralized finance (DeFi) protocols. Rocky independently makes decisions without reliance on a central authority. Notably, the bot leverages Zero-Knowledge Machine Learning (ZKML) for AI analysis on user biometric data, ensuring data custody.

Through the use of zero-knowledge cryptography, proofs of Rocky’s trading model — including his decision-making processes, model inputs, dietary patterns, and emotional states — are systematically recorded and verified on the Ethereum blockchain. This capability is facilitated by StarkNet, an advanced Layer 2 roll-up solution developed by Starkware.

  • Uphshot — decentralized machine intelligence network.

Upshot’s core mission centers around facilitating the establishment of efficient financial markets across diverse domains, underpinned by our proprietary Proof of Alpha consensus mechanism. This mechanism systematically consolidates and incentivizes participants within the network based on the efficacy of their contributions, aligning with the optimization of specific objective functions.

The platform enhances security through verifiable computation protocols, crucial for network-integrated applications. This involves implementing a novel zkSNARK proof system designed to validate tree-based machine learning model outputs. A notable application is zkPredictor, a cutting-edge tool developed in collaboration with and powered by Modulus Labs. It utilizes machine learning models to predict asset prices, representing a significant advancement in decentralized, self-improving intelligence networks for financial market modeling.

World ID is an advanced digital passport leveraging ZKML for AI analysis on user biometric data, ensuring secure custody of information. This protocol enables users to substantiate their unique and authentic identity while preserving anonymity, achieved through the utilization of zero-knowledge proofs (ZKP) and other privacy-centric cryptographic mechanisms.

Integration of World ID into applications, smart contracts, or backend systems is seamless. Holders of World ID can effortlessly and anonymously confirm their distinct human identity through a compatible identity wallet, such as the World App.

Fundraising Rounds and Grant Programs for ZKML Initiatives

In the preceding year, numerous venture capital funds and private investors within the cryptocurrency market demonstrated confidence in Zero-Knowledge Machine Learning (ZKML) technology, making substantial financial investments in this innovative domain.

Total fund raised by ZKML initiatives

The diagram elucidates a substantial influx of investments in the ZKML technology-centric projects from prominent venture funds, including but not limited to Polychain Capital, Blockchain Capital, Delphi Digital, Framework Ventures, Jump Capital, Galaxy, 1kx, Variant, and Stanford University.

Polychain Capital emerged as the most prolific fund, engaging in 5 distinct investments amounting to an aggregate sum exceeding $11 million.

Grant programs also actively support the integration of ZKML technology, incentivizing developers to generate inventive concepts for incorporating ZeroKnowledge into Machine Learning and fortifying established products like WorldCoin as an example.

Worldcoin is offering targeted grant opportunities and inviting applications for up to 25,000 $WLD tokens in two primary domains:

  1. Iris Code Zero-Knowledge Machine Learning (zkML): This encompasses the calculation of iris codes from images utilizing Zero-Knowledge proof, facilitating self-custodial advancements to iris codes.
  2. Face Authentication Zero-Knowledge Machine Learning (zkML): This involves the computation of face embeddings and their comparison with the Orb’s embeddings, thereby reducing trust assumptions on the client side.

Potential use cases and applications for ZKML technology

  • Zero-Knowledge Anomaly/Fraud Detection:

Facilitates the development of a Zero-Knowledge proof for assessing exploitability or fraudulent activities. Anomaly detection models, trained on smart contract data and validated by DAOs, can serve as significant metrics for automating security protocols, allowing for a more proactive and preventive approach, such as pausing contracts. Given the current exploration of machine learning models for security within the smart contract realm, the implementation of Zero-Knowledge anomaly detection proofs represents a potential progression in this domain.

  • Transparency in Machine Learning as a Service (MLaaS):

In instances where various companies offer access to Machine Learning models via their APIs, determining the veracity of the model being provided can be challenging for users, given the black-box nature of APIs. Supplying validity proofs linked to an ML model API becomes instrumental in offering transparency to users, allowing them to verify the authenticity of the utilized model.

  • Privacy in Zero-Knowledge Machine Learning (ZKML):

In addition to validity proofs, the capacity to conceal specific aspects of the computation is available to facilitate the privacy-preserving deployment of Machine Learning.


The ZKML community, established in 2023, comprises various teams and individuals actively involved in advancing Zero-Knowledge Machine Learning.

Numerous organizations are committed to enhancing ZK technology by developing optimized hardware to accelerate the computation of Zero-Knowledge proofs, particularly for resource-intensive operations like prover and verifier algorithms. With advancements in specialized hardware, improvements in proof system design (proof size, verification time, proof generation time, etc.), and more efficient implementations of ZK protocols, the ability to demonstrate larger models on less powerful computers in a shorter timeframe is expected as ZK technology progresses.

While ZKML is in its early developmental stage, it has attracted significant investor interest, already demonstrating substantial value. The ongoing evolution of ZKML suggests the emergence of innovative applications, pointing towards a future where privacy-preserving machine learning becomes the norm.

Schedule a call with our experts:

About us

Web3 Architects by Gotbit is a multidisciplinary team of over 150 blockchain analysts, protocol architects, token economists, and web3 venture experts. Since 2017, we’ve assisted more than 500 projects in designing tokens, modeling protocol economics and incentives, attracting users and liquidity, and securing investments from top funds. We’ve effectively guided hundreds of projects from token design to post-launch support.