Project Overview

GhostProver is a privacy-preserving AI compliance layer that allows teams to prove sensitive data was not sent to AI models without exposing the original prompt. Instead of storing raw prompts or relying on trust, GhostProver combines Zero-Knowledge proofs, TEE-verified inference, decentralized storage, and on-chain receipts into a verifiable compliance pipeline.

Designed for AI workflows in fintech, healthcare, KYC, enterprise systems, and AI agents.

Problem

Current AI compliance solutions have major limitations:

• Prompt logging creates privacy risks

• DLP systems detect issues but provide no cryptographic proof

• Audits rely heavily on trust

GhostProver enables privacy-preserving proof of compliance.

Core Architecture

Key Features

• Noir ZK circuit for non-inclusion proofs

• Poseidon2 commitments for prompt privacy

• TypeScript SDK, CLI, Batch Prover, and MCP server

• React operator dashboard

• Policy packs for SaaS, KYC, Banking, Fintech, and Healthcare

• Batch compliance verification across multiple sensitive-data patterns

Main role in this project was of 0G ecosystem.

0G Integration

0G Chain

Deployed on mainnet:

• HonkVerifier contract: 0x17b9d7b36bf6e77f7dbc010b4b2be662a3f1df78

• GhostProverRegistry contract: 0x9595BD4e6b868C64001904EeF76d838D78604B6e

Generates on-chain compliance receipts containing proof and audit metadata.

0G Compute

Used for live inference, provider discovery, and TEE verification.

0G Storage

Stores decentralized audit evidence and proof bundles.

Mainnet Results

• Full end-to-end run on 0G Mainnet

• Verified TEE inference

• 9/9 SaaS compliance proofs

• 0G Storage archival

• On-chain batch compliance receipts

Impact

GhostProver turns AI compliance from a trust-based claim into a verifiable cryptographic workflow.

During the hackathon, GhostProver evolved from an idea into a working end-to-end privacy-preserving AI compliance system that proves sensitive information was not included in AI prompts without exposing the prompt itself.

Zero-Knowledge Compliance Layer

Built a Noir-based ZK non-inclusion circuit using Poseidon2 commitments to prove prompts do not contain sensitive patterns while keeping prompts private.

Implemented:

• Exact and pattern-based checks

• Support for API keys, Aadhaar, PAN, JWTs, credit cards, bearer tokens, database URLs, etc.

• 17 test cases covering valid proofs, failures, tampering, and edge cases

Developer & Product Infrastructure

Built the full developer workflow:

• TypeScript SDK + CLI

• Batch prover for multi-policy verification

• Express middleware

• Daemon API

• MCP server for AI agents

• React operator console

Policy System

Created policy registries and presets for:

• SaaS

• KYC

• Banking

• Fintech

• Healthcare

Added support for custom compliance rules and sensitive-data pattern libraries.

0G Integration

Integrated core 0G infrastructure:

0G Compute

• Live mainnet inference

• Provider discovery

• Request authentication

• TEE verification

0G Storage

• Audit bundle archival

• Evidence storage and storage root generation

0G Chain

Deployed on mainnet:

• HonkVerifier contract: 0x17b9d7b36bf6e77f7dbc010b4b2be662a3f1df78

• GhostProverRegistry contract: 0x9595BD4e6b868C64001904EeF76d838D78604B6e

Reliability Improvements

During the hackathon we also:

• Fixed 0G SDK mainnet issues

• Added network-aware configuration

• Improved runtime and environment handling

• Strengthened validation and frontend/API flows

Outcome

GhostProver now combines Zero-Knowledge compliance proofs, TEE-verified inference, decentralized storage, on-chain receipts, and AI-agent integrations into a working verifiable AI compliance system.

GhostProver is at a pre seed stage. It has moved beyond a hackathon prototype into a working end to end system across ZK proofs, TEE-verified inference, 0G Storage, and 0G Chain.

During the hackathon, we built the core stack: Noir ZK circuits, Poseidon2 commitments, batch proof generation, TypeScript SDK, CLI, MCP server, daemon API, React operator console, 0G Compute integration, audit bundle uploads to 0G Storage, and on chain compliance receipts. We also deployed HonkVerifier and GhostProverRegistry on 0G mainnet and completed a live mainnet run with verified inference and 9/9 SaaS compliance proofs.

We are not raising a large institutional round yet. The immediate goal is a small pre seed round, ecosystem grant, or strategic angel round to fund audits, product hardening, enterprise workflows, policy pack expansion, go to market, and early pilots.

Our strongest fundraising asset is technical credibility: deployed contracts, real mainnet receipts, and a clear market need. GhostProver is building the privacy preserving compliance layer for AI inference, letting teams prove sensitive data was not sent to AI systems without exposing the data itself.