◎ OS PUB Apache 2.0 ← All specifications

P247 — AIEP — Reasoning Receipt and Actuation Proof Format

Applicant: Neil Grassby Classification: Patent Application — Confidential Priority: Claims priority from GB2519711.2 filed 20 November 2025 Architecture Layer: AIEP Phase 2 Audit and Proof Layer

Framework Context

[0001] This specification operates within an AIEP environment as defined in GB2519711.2 and GB2519798.9. The present specification defines a standardised proof format — the Reasoning Receipt — that is produced for each reasoning session and actuation event, providing a compact verifiable record linking each action taken to the reasoning chain and evidence that authorised it.

Field of the Invention

[0002] The present invention relates to verifiable reasoning receipt formats for action auditability in evidence-bound AI systems.

Background

[0003] External oversight of AI system actions requires the ability to verify post-hoc that: a specific action was taken; the action was authorised by a valid reasoning process; the reasoning process was grounded in admitted evidence; and governance constraints were satisfied. The Reasoning Receipt format provides this verifiable link in a compact, standardised form.

Summary of the Invention

[0004] The invention provides a Reasoning Receipt and Actuation Proof Format (RRAPF) that defines the structure and content of Reasoning Receipts. Each receipt records: action specification and timestamp; governance decision certificate (P242); evidence activation set Merkle root; reasoning session identifier and checkpoint hash (P245); simulation certification hash (P233) if applicable; and the system identity signature over all fields. Receipts are admitted to the evidence ledger and discoverable by any authorised auditor.

ASCII Architecture

Action Execution Event (P206)
         |
         v
+------------------------------------------+
| Reasoning Receipt and Actuation Proof    |
|   Format Engine (RRAPF)                  |
|                                          |
|  Collect:                               |
|    - Action specification + timestamp   |
|    - Governance certificate (P242)      |
|    - Evidence activation set hash       |
|    - Session checkpoint hash (P245)     |
|    - Simulation certificate (P233)      |
|  Construct Reasoning Receipt record     |
|  Sign with system identity key          |
|  Admit to evidence ledger               |
+-------------------+----------------------+
                    |
                    v
         Reasoning Receipt → Ledger
         → Discoverable by auditors

Detailed Description

[0005] Receipt Structure. The Reasoning Receipt is a JSON-serialisable record with the following top-level fields: receipt_id (UUID), action_spec (hash + description), action_timestamp, governance_cert_hash, evidence_activation_root (Merkle root), reasoning_session_id, session_checkpoint_hash, simulation_cert_hash (null if no simulation), system_signature.

[0006] Ledger Admission. The completed receipt is admitted to the evidence ledger as a first-class receipt artefact, carrying ledger sequence number and Merkle chain membership. The ledger admission provides independent non-repudiation of the action event.

[0007] Audit Query. Authorised auditors may query the evidence ledger for receipts by action type, timestamp range, goal identifier, or governance policy version. The audit query API returns receipt records with their full ledger membership proofs.

[0008] Verification. A verifier confirms a receipt by: checking the system signature; verifying the governance certificate against the enclave public key; confirming the evidence activation root is consistent with archived evidence artefacts; and optionally replaying the reasoning session from the checkpoint hash.

Technical Effect

[0009] The invention provides a machine-verifiable cryptographic link between each AI system action and the complete evidentiary, reasoning, and governance chain that authorised it, enabling post-hoc accountability without access to live system state. By binding the governance certificate hash, evidence activation Merkle root, session checkpoint hash, and simulation certificate hash in a single system-signed receipt admitted to the evidence ledger, the format enables any authorised verifier to independently confirm the evidentiary basis of any recorded action. The audit query API with ledger membership proofs enables compliant evidence retrieval for regulatory investigation.

Claims

  1. A method of generating and admitting a reasoning receipt actuation proof in an evidence-bound artificial intelligence system, comprising the steps of: (a) constructing a reasoning receipt record containing: a UUID receipt_id; an action_spec hash and description; an action_timestamp; a governance_cert_hash referencing the authorising governance decision certificate; an evidence_activation_root as the Merkle root over the evidence activation set; a reasoning_session_id; a session_checkpoint_hash; and a simulation_cert_hash if simulation was performed; (b) signing the receipt record with the system’s private identity key, producing a system_signature field; (c) admitting the signed receipt to the evidence ledger as a first-class receipt artefact, receiving a ledger sequence number and Merkle chain membership; (d) providing an audit query API enabling retrieval of receipts by action type, timestamp range, goal identifier, or governance policy version, each query response including ledger membership proofs.

  2. The method of claim 1, wherein the evidence_activation_root is computed as the Merkle root of all evidence artefact identifiers in the activation set at the time of the action.

  3. The method of claim 1, wherein verification of a receipt comprises: checking the system_signature; verifying the governance_cert_hash against the governance execution enclave’s public key; confirming the evidence_activation_root against archived evidence artefacts; and optionally replaying the session from the session_checkpoint_hash.

  4. The method of claim 1, wherein audit queries with governance policy version filters return all receipts produced under the specified policy version, enabling audit scope definition by regulatory period.

  5. The method of claim 1, wherein simulation_cert_hash is a mandatory non-null field when the authorising reasoning included any counterfactual simulation step.

  6. A reasoning receipt generation and audit system for an evidence-bound artificial intelligence system, comprising: a receipt constructor building structured JSON receipt records with governance, evidence, and checkpoint hash fields; a signing module applying the system private key; a ledger admission module; and an audit query API with membership proof generation.

  7. A computer-readable medium carrying instructions for implementing the method of any preceding method claim.

Abstract

A reasoning receipt actuation proof format for evidence-bound artificial intelligence produces a system-signed JSON receipt record for each AI system action, containing a UUID, action specification hash, governance decision certificate hash, Merkle root over the evidence activation set, session checkpoint hash, and simulation certificate hash. The signed receipt is admitted to the evidence ledger as a first-class artefact. An audit query API with ledger membership proofs enables authorised parties to retrieve receipts and independently verify the evidential, governance, and reasoning chain underpinning any recorded action.

Dependencies

This specification is published as open-source prior art under the Apache License 2.0. Copyright 2026 Protocol Owner. Inventor: Protocol Architect.

Patent information · Licensing · All open specifications