P237 — AIEP — Knowledge Reconstitution Proof Engine

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

Framework Context

[0001] This specification operates within an AIEP environment as defined in GB2519711.2 and GB2519798.9. The present specification defines a mechanism for generating and verifying cryptographic proofs that a distilled or compressed knowledge claim can be fully reconstituted from its original evidence artefacts.

Field of the Invention

[0002] The present invention relates to knowledge reconstitution proof systems for compressed knowledge in evidence-bound artificial intelligence.

Background

[0003] The Knowledge Distillation Engine (P217) compresses large evidence corpora into compact distilled claims. These claims are used in reasoning but their connection to original evidence may weaken over time as compression artefacts accumulate. A reconstitution proof mechanism provides a cryptographic guarantee that the full evidential basis of a distilled claim is recoverable.

Summary of the Invention

[0004] The invention provides a Knowledge Reconstitution Proof Engine (KRPE) that, at distillation time, computes a reconstitution proof for each distilled claim. The proof consists of: the Merkle root of all source artefact hashes; the distillation parameters applied; a hash of the distilled claim; and a signed commitment that the distillation function applied to the source artefacts produces the stated distilled claim.

[0005] Reconstitution verification is performed by: re-applying the distillation function to the source artefacts; comparing the output hash to the certified claim hash in the proof record. A mismatch indicates artefact corruption, parameter drift, or tampered claims.

ASCII Architecture

Knowledge Distillation (P217)
         |
         v
+------------------------------------------+
| Knowledge Reconstitution Proof Engine    |
|   (KRPE)                                 |
|                                          |
|  Source artefact hash collection        |
|  Merkle root construction               |
|  Distillation parameter recording       |
|  Proof record construction + signing    |
|  Proof record → Evidence Ledger         |
+-------------------+----------------------+
                    |
         Reconstitution verification
         - Replay distillation
         - Compare output hash to proof

Detailed Description

[0006] Proof Construction. At distillation time, the KRPE receives the source artefact set used by the distillation run. It constructs a Merkle tree over the source artefact hashes, producing a Merkle root that commits to the full source set without requiring all artefacts to be stored in the proof.

[0007] Commitment Signature. The KRPE signs the proof record with the system identity key, preventing post-hoc modification of the reconstitution parameters or source set.

[0008] Ledger Admission. The signed proof is admitted to the evidence ledger as a distillation certificate. This makes the proof discoverable and auditable by any party with ledger access.

[0009] Reconstitution Verification. To verify a distilled claim, the verifier: retrieves the source artefacts referenced in the Merkle tree; re-applies the recorded distillation parameters to the source set; produces an output hash; compares it to the claim hash in the proof record. Verification success confirms the claim is faithfully derived from its source evidence.

[0010] Staleness Detection. If any source artefact is subsequently modified or retracted, the Merkle root becomes invalid. The proof engine detects this during periodic proof audit cycles and flags affected distilled claims for re-distillation.

Technical Effect

[0011] The invention provides cryptographic verifiability of the distillation chain linking distilled knowledge claims to their source evidence artefacts. By constructing a Merkle tree over the source artefact set and recording distillation parameters alongside the root hash, the engine enables any verifier to confirm that a distilled claim is faithfully derived from its claimed sources by deterministic replay. By detecting source artefact retraction through periodic Merkle root revalidation, the engine ensures that distilled claim integrity is continuously maintained after issuance.

Claims

1. A computer-implemented method for knowledge reconstitution proof, the method comprising: (a) receiving a distilled claim record from the Knowledge Distillation Engine and retrieving the source evidence artefacts referenced in its provenance chain; (b) constructing a Merkle tree over the source artefact set and recording the root hash and individual leaf hashes; (c) recording the distillation parameters applied to produce the distilled claim; (d) generating a reconstitution proof record comprising the Merkle root, distillation parameters, distilled claim hash, and system signature; (e) admitting the signed proof record to the AIEP evidence ledger as a distillation certificate; and (f) performing periodic proof audit cycles to verify that source artefact hashes in the Merkle tree remain consistent with ledger records, flagging affected distilled claims for re-distillation on retraction detection.

2. The method of claim 1, wherein reconstitution verification is performed by: retrieving source artefacts referenced in the Merkle tree; re-applying recorded distillation parameters; computing the output hash; and comparing it to the claim hash in the proof record.

3. The method of claim 1, wherein the Merkle tree construction order over source artefacts is deterministic, enabling identical tree construction on independent verification.

4. The method of claim 1, wherein distillation certificates admitted to the ledger are cross-referenced to their associated distilled claim records, enabling bilateral traversal of the proof relationship.

5. The method of claim 1, wherein proof audit cycle frequency is configurable via the active governance policy.

6. A Knowledge Reconstitution Proof Engine comprising: one or more processors; memory storing a Merkle tree builder, distillation certificate store, and proof audit scheduler; wherein the processors are configured to execute the method of claim 1.

7. A non-transitory computer-readable medium storing instructions that, when executed by a processor, implement the method of claim 1.

Abstract

A knowledge reconstitution proof engine for evidence-bound artificial intelligence generates cryptographic proofs of the distillation chain linking distilled knowledge claims to their source evidence artefacts, using a Merkle tree over the source set and recorded distillation parameters. Signed proof records are admitted to the AIEP evidence ledger as distillation certificates. Periodic audit cycles detect source artefact retractions by revalidating Merkle roots, flagging affected claims for re-distillation to maintain continuous knowledge integrity.