HOW IT WORKS

From AI output to verifiable proof

Four steps. No complexity. Verify in your browser

01

Your AI generates a response

A customer asks a question. Your AI answers. This interaction — the exact input, output, timestamp, and model — is the content you want to prove.

02

Content is hashed

A SHA-256 hash is computed from the interaction. This creates a unique fingerprint — 64 hexadecimal characters. Any change to the content, even a single space, produces a completely different hash.

03

Hash is sealed in a chain

The hash is inserted into a Merkle tree and included in a checkpoint. The checkpoint is signed with an Ed25519 key and linked to the previous checkpoint via its hash. This creates an append-only chain where any modification is detectable.

04

Proof is verifiable in your browser

Anyone can verify a proof by recomputing the hash from the original content and comparing it to the stored hash. Verification happens in the browser against Sealison's pinned public key — no live server call needed.

TAMPER EVIDENCE

Why this is tamper-evident

Hash binding

Each record is bound to its content via SHA-256. Change one byte, the hash changes completely.

Chain linking

Each checkpoint includes the hash of the previous checkpoint. Altering any record breaks the chain.

Signature sealing

Each checkpoint is signed with Ed25519. The signature proves the chain state at that moment in time.

Client-side verification

Anyone can recompute the hash and verify the signature against Sealison's pinned public key, client-side.

HASH-ONLY MODE

Your content stays on your system

In hash-only mode, the server does not receive the original content. Only the SHA-256 hash is transmitted and sealed. The data owner retains full control of their content while still getting a verifiable proof of integrity.

This is the recommended mode for enterprise use. It provides cryptographic evidence supporting auditability and selected record-keeping workflows.

EU AI ACT · REGULATION 2024/1689

Supporting auditability by design.

Article 12 requires AI systems to maintain automatic, tamper-evident logs. Standard application logs can be useful operational records, but they are usually not designed as standalone cryptographic proof records.

SEALISON provides cryptographic tamper-evidence by design. Each record is hashed, signed, and chained. Any modification is detectable by re-running the verifier against Sealison's pinned public key, client-side.

Ready to try it?

▲ Quickstart →

Logs are claims.
Proof is verifiable.

HomeVerifyRetailFintechHealthcareDocsAPIQuickstart
LegalPrivacyCookies
SealisonProof, not logs