HOW IT WORKS

From AI output to verifiable proof.

Four steps. No complexity. No trust required.

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 forever

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 — no server trust required.

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.

Independent verification

Anyone can recompute the hash and verify the signature. No need to trust the server.

HASH-ONLY MODE

Your content never leaves your system.

In hash-only mode, the server never sees 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 satisfies EU AI Act Article 12 requirements for tamper-evident logging without requiring data to leave the customer’s infrastructure.

EU AI ACT · REGULATION 2024/1689

Compliance by design.

Article 12 requires AI systems to maintain automatic, tamper-evident logs. Standard application logs (ELK, Datadog, CloudWatch) do not meet this requirement — they can be modified without detection.

SEALISON provides cryptographic tamper-evidence by design. Each record is hashed, signed, and chained. Any modification is detectable by anyone, at any time, without server access.

Ready to try it?

▲ Quickstart →

Verifiable infrastructure
for AI systems.

HomeVerifyRetailFintechHealthcareDocsAPIQuickstart
LegalPrivacyCookies
SEALISON · Powered by Immutal