TruthAnchor — Evidence & Verification for Outcome-Based AI Work
Live registry stats: 743 total proofs · 743 verified · 0 pending. Anchored across Polygon, Base, BNB Smart Chain mainnets plus Sepolia testnet. Figures reflect the registry database at the time this page was rendered.
Know when AI work is actually complete — before settlement. TruthAnchor creates independently verifiable evidence records, verification results, and clearing receipts for outcome-based AI work. Non-custodial: payment stays on your existing rails. Built on open, on-chain anchoring (ERC-8263).
Who decides when an AI task is actually complete?
When one economic party pays another for AI work, and payment depends on whether the agreed outcome was actually delivered, evidence controlled by one of the parties may not be sufficient for both sides in a dispute. TruthAnchor sits between the parties: independently verifiable evidence records, verification of agreed success criteria, and clearing receipts. TruthAnchor is most useful when payer and performer are different economic parties and settlement depends on an objectively verifiable outcome. If the payer and performer are the same entity, existing internal logs may already be sufficient. TruthAnchor does not custody customer funds, operate escrow, or move money — settlement stays on external payment rails chosen by the parties. Start with a Manual Clearing Pilot at /pilot.
Four Capability Lines, One Trust Stack
Anchor: every agent action is hashed and sealed across multi-chain (Polygon, Base, BSC mainnets + Sepolia testnet). The original data stays private — only the proof is public. Verify: trace any output back to its source data and grounding evidence. Independent verification, no account required. Reputation: a reputation score computed from an agent's anchored on-chain history — evidence-derived, not self-reported, and verifiable without exposing the underlying data. Audit Trail: anchored actions can include source, session and parent-transaction references, laying the groundwork for causal-chain reconstruction. Full causal replay and compliance export are in development.
How It Works
1. Register — AI agents register with a unique identity. Every action is attributed to a known agent. 2. Anchor — actions, data sources, and outputs are hashed and anchored on-chain. Only the SHA-256 hash is stored. Your data stays private. 3. Verify — anyone can verify any proof independently. Trace the full chain: data source → decision → action. No account required.
Multi-Chain Coverage
Product anchoring runs on Polygon, Base, and BNB Smart Chain mainnets plus Sepolia testnet. The verified ERC-8263 reference contract is on Ethereum Mainnet — used for standard credibility and verification, not as the default user anchor path.
Built for AI at Scale
Trading Bots: every trade decision anchored on-chain with the market data source that triggered it. Clients verify your bot's actions independently. Audit & Compliance: tamper-evident agent logs for regulated industries (finance, healthcare, legal). Full decision replay and compliance export are in development. Research Reproducibility: anchor model outputs, training data hashes, and eval scores. Anyone can verify the chain of evidence behind a published result.
Open Standard — ERC-8263
TruthAnchor is the reference implementation of ERC-8263 (Onchain Proof Layer for AI Agent Actions). The reference contract is verified on Ethereum Mainnet at 0xe95d6a15966984c209a62a2c188828555eb5ec3d. Read the full spec at /spec.