Verifiable receipts · live on Base

Every API call, provable.

NextModel produces a verifiable receipt for every billable call, hashes them into Merkle batches, writes the archives with retention locks, and anchors each batch on the Base blockchain. A tamper-evident AI audit trail you do not have to take our word for.

Live proof

Loading the latest anchor batch…

How it works

From one call to an on-chain anchor in four steps.

1 · Receipt hashed

Every billable request produces a structured receipt — model, provider, tokens, cost, timing — hashed with SHA-256 at write time.

2 · Merkle batch

Receipt hashes are grouped into Merkle trees, so a single 32-byte batch root commits to every receipt in the batch.

3 · Retention-locked archive

Each batch manifest is written to object storage under a retention lock — write-once, so the archived record cannot be quietly rewritten.

4 · Anchored on Base

The batch Merkle root is anchored on Base, an Ethereum L2 — an independent, public timestamp anyone can inspect in a block explorer.

Check one yourself, right now.

If you hold a NextModel receipt, the receipt verifier tool checks it against the anchored batches and returns its Merkle membership proof. No login is required for the live proof panel above; receipt verification asks only for what the receipt already shows.

Open the receipt verifier

Want the full provenance background? Read the AI provenance guide.

FAQ

Verifiable AI API questions

What is a verifiable AI API?

A verifiable AI API pairs every response with evidence you can check independently. NextModel pairs each billable call with a receipt, hashes receipts into Merkle batches, locks the archives with retention policies, and anchors each batch root on the Base blockchain, so claims about what happened can be verified instead of trusted.

What does a NextModel receipt contain?

A receipt records the model that served the request, the provider behind it, token counts, cost, timing, and a hash of the request. Receipts are structured, small, and designed so a third party can later confirm what happened without access to your prompt content.

Where are the receipts anchored?

Receipt hashes are grouped into Merkle batches. Each batch manifest is written to retention-locked, write-once object storage, and the batch's Merkle root is anchored on Base, an Ethereum layer 2. Anchors are ordinary transactions you can open on a block explorer.

Can I verify a receipt myself?

Yes. The receipt verifier tool checks a receipt and returns its Merkle membership proof, and the live proof panel on this page shows the latest anchored batch with a link to its on-chain transaction.

Does verification expose my request content?

No. Anchoring covers hashes and batch metadata, not payloads. Prompts and responses stay private; what becomes publicly checkable is that a receipt existed, what it committed to, and that it was not altered afterward.

Is anchoring a future plan or shipped?

Shipped. The production loop has been live since September 2026: usage is scanned on a schedule, receipts are batched and archived with retention locks, and batch roots are anchored on Base. The panel above reads the live pipeline.