Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.blindference.xyz/llms.txt

Use this file to discover all available pages before exploring further.

​
Welcome to Blindference

Blindference is a confidential AI execution layer for Web3. It enables users to submit sensitive prompts and data to frontier AI models — Groq Llama 3, Google Gemini — without ever revealing them to any single party.

​
What It Does

  • Input Privacy: User prompts are AES-256 encrypted in the browser before leaving the device
  • Access Control: Encryption keys are split into FHE-encrypted halves and stored on-chain via CoFHE, only decryptable by assigned quorum nodes
  • Execution Integrity: A 1 leader + 2 verifier quorum runs identical inference and cross-validates results
  • Economic Accountability: Disputed results trigger on-chain verification with automatic USDC payouts via Reineira
  • Output Privacy: Only the user can decrypt the final result using their wallet

​
Who Is This For?

Compute Providers

Run a Blindference node to earn fees executing encrypted inference jobs. GPU optional.

Agent Developers

Build agents that use Blindference for confidential inference, risk scoring, or custom AI workflows.

​
Supported Modes

  • Confidential Text Inference: Submit natural language prompts through encrypted channels
  • Confidential Risk Scoring: Submit structured financial features for privacy-preserving evaluation

​
System Architecture

Run a Node

Start earning in 5 minutes

Build an Agent

Integrate confidential inference

Contract Addresses

Arbitrum Sepolia deployments