A cryptographic AI receipt is signed, independently verifiable proof that an AI event happened the way it is claimed. Hive mints receipts for AI inference, agent actions, and human-origin data. Each one is signed with post-quantum ML-DSA-65 (NIST FIPS 204) and verifiable offline by any third party, without trusting the platform that produced them. Start at the AI receipts hub for the full proof layer.
What they are
A receipt captures the conditions of an AI event, meaning the model, inputs, decisions, and timing, and binds them into a signed record. The signature is cryptographic: anyone holding the public key can confirm the record was produced by the claimed signer and has not been altered. It is independent: verification runs offline and needs nothing from Hive at check time. And it is honest about scope. A receipt attests what happened, never a route on whether an output is "correct."
The three surfaces
Every inference, and each sub-step of it, is cryptographically attested with ML-DSA-65, so a signed model call can be verified after the fact by anyone.
Explore AFiRAgent receipts turn route decisions, tool calls, and model choices into a signed receipt chain: verifiable proof of what an autonomous agent actually did.
See agent receiptsOriginProof makes expert-data provenance buyer-verifiable: signed proof that data was human-conditioned, checkable without trusting the platform's own assertion.
See OriginProofAssurance
Different tasks need different amounts of proof. Hive's assurance tiers map the same receipt primitive to signed inference, verifiable agents, and regulated workflows, so you pick the level of independent third-party proof a task requires. Verification is always free; you only pay to sign.
Paid receipts
When a receipt carries a price, the payment rail is x402: a signed USDC quote on Base, settled machine-to-machine. Verifying a receipt stays free; paying to mint one is a real, self-serve path with no checkout theater.