What Is Dead Man's Proof?
A privacy-preserving attestation system that lets you prove things about private data without ever revealing it.
The problem with verification today
Every time you need to prove something about yourself, you hand over the full picture. Want to prove you have five years of engineering experience? Share your entire resume. Want to prove you hold a certain asset balance? Reveal your whole portfolio. Want to confirm you're free on Thursday? Expose your entire calendar.
Verification today is built on a broken assumption: that proving a claim requires disclosing the underlying data. This forces people to choose between privacy and credibility.
Dead Man's Proof eliminates that tradeoff.
How it works
Seal your data into the vault
Submit private information, whether that's a resume, financial records, or a calendar. The system computes a keccak256 hash of your data and publishes that hash commitment onchain to Base.
Your raw data is stored securely and never exposed. The onchain hash serves as a tamper-proof anchor: if anyone tries to alter the sealed data, the hash won't match.
Ask yes or no questions
Anyone with access to a vault's commitment ID can submit natural language yes/no questions. "Does this person have 5+ years of experience?" "Does the depositor hold more than 10 ETH?" "Is the person available on Tuesday afternoon?"
Questions are constrained to yes/no format by design. This prevents data extraction through open-ended queries and ensures the underlying data stays sealed.
Cryptographic attestation, no secrets revealed
An AI agent evaluates the query against the sealed data using privacy-first inference with no data retention. The result is a cryptographic attestation published onchain, containing:
- ✞Verdict: YES or NO
- ✞Confidence: 0 to 100% certainty score
- ✞Reasoning: Privacy-safe explanation that never leaks the underlying data
- ✞Onchain record: Permanently stored on Base with the commitment hash
What you can prove without revealing
Hiring Verification
Candidates seal their resume. Recruiters verify specific claims ("Do they have 5+ years in React?") without accessing the full document. No more oversharing in job applications.
Financial Proof
Users seal portfolio data. Counterparties verify holdings ("Do they hold more than $50k in stablecoins?") without seeing exact balances. Prove solvency without exposure.
Availability Checking
Users seal their calendar. Others check availability ("Are they free Thursday 2pm to 4pm?") without seeing the full schedule. Coordination without surveillance.
Compliance Attestation
Seal sensitive documents. Auditors verify compliance claims without accessing raw data. Meet requirements without handing over the keys.
Why Venice AI?
Privacy is the product. The inference layer cannot store, log, or train on vault data. Venice AI enforces zero data retention at the infrastructure level: sealed data enters the model, the attestation comes out, and nothing persists.
The agent reasons over confidential information and then acts on it publicly. It evaluates private resumes, financial records, and calendars, then publishes verifiable attestations onchain. Venice makes this possible because the sensitive data never leaves the session boundary.
No other inference provider offers this guarantee at the protocol level. Without it, a privacy-preserving attestation system would be privacy theater.
Why Base?
Attestations need to be cheap, fast, and permanent. Base provides low transaction costs so every attestation can be committed onchain without friction. The Ethereum security model ensures that once an attestation is published, it cannot be altered or removed.
The DeadMansVault smart contract stores hash commitments and attestation records with agent-only access control. Only the authorized AI agent can publish attestations, preventing unauthorized verdicts from entering the record.
Security model
Privacy is the product, so the security model is built to match:
- ✞Privacy-first inference: AI evaluations run with no data retention. The inference provider never stores or trains on vault data.
- ✞Prompt injection defense: Five layers of protection including input sanitization, message isolation, hardened system prompts, leak detection, and output truncation.
- ✞Rate limiting: 5 queries per minute per IP and 200 per day globally to prevent abuse.
- ✞Onchain access control: Only the authorized agent wallet can publish attestations to the contract.
Ready to seal your data?
Your secrets enter the vault. Only the truth comes out.