◎ OS PUB Apache 2.0 ← All specifications

P134 — AIEP — Autonomous Mirror Generation Protocol

Publication Date: 2026-03-15 Status: Open Source Prior Art Disclosure Licence: Apache License 2.0 Author/Organisation: Phatfella Ltd Schema: AIEP_OS_SPEC_TEMPLATE v1.0.1 — https://aiep.dev/schemas/aiep-os-spec-template/v1.0.1

Framework Context

[0001] This disclosure operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, and GB2519803.7, filed 20 November 2025, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure extends the mirror generation mechanisms defined in P60–P64 and P132 to the autonomous creation of verifiable evidence mirrors by AIEP-governed systems during evidence retrieval operations, without requiring human curation of the source set.

Field of the Disclosure

[0003] This disclosure relates to autonomous evidence mirror generation in governed artificial intelligence systems.

[0004] More particularly, the disclosure concerns a protocol under which an AIEP-governed system automatically detects missing evidence mirrors during retrieval operations, extracts and normalises content from the source, computes a cryptographic content hash, creates a structured MirrorRecord, and registers the record in the Distributed Evidence Index, expanding the verifiable evidence network without manual intervention.

Background

[0005] Governed AI reasoning systems require that every ingested evidence artefact reference a stable, verifiable source. Web content is inherently unstable: URLs change, content is updated, and sources are removed. Without a stable structured mirror, an evidence artefact’s content hash cannot be reliably verified after the fact.

[0006] Manual mirror curation — in which an operator selects sources to mirror and triggers the mirror creation process — does not scale to the volume of sources that a continuously operating AIEP reasoning system discovers and ingests.

[0007] No existing specification defines a normative trigger condition, extraction pipeline, MirrorRecord schema, and index registration procedure that enables a governed AI system to create and register structured evidence mirrors autonomously during its normal retrieval operations.

Summary of the Disclosure

[0008] Mirror Generation Trigger. A mirror generation event is initiated when: (a) evidence retrieval identifies a source URL with no existing MirrorRecord in the local or federated evidence index; (b) the source’s HTTP fetch status is successful; and (c) the source passes the governance admissibility gate (power budget, payload size, domain credibility checks).

[0009] Content Extraction. The retrieved content body is processed through a metadata normalisation pipeline extracting: document title; author or publishing organisation (if determinable from structured metadata); publication date (if present); jurisdiction (inferred from domain TLD, content language, and explicit metadata); and topic classification tags from the schema-defined taxonomy.

[0010] Content Hashing. A content_hash is computed as SHA-256 over the raw retrieved content body prior to normalisation. This hash binds the MirrorRecord to the specific content state at extraction time, enabling tamper detection on future retrieval.

[0011] A MirrorRecord is constructed comprising: mirror_id derived as the SHA-256 hash of source_url concatenated with extraction_timestamp; source_url; mirror_url identifying the durable storage location; content_hash; extraction_timestamp; title; author; publication_date; jurisdiction; classification_tags; and canonical_source_reference.

[0012] Mirror Persistence. The MirrorRecord and a content snapshot are persisted to the node’s configured durable storage (R2 object store or equivalent), with the content snapshot stored under the key mirrors/{sha256_prefix}/{mirror_id}.json.

[0013] Index Registration. An EvidenceIndexEntry is derived from the MirrorRecord and submitted to the node’s local index partition. The entry is propagated to federated index partitions according to the distribution rules defined in P133.

[0014] Validation. Before acceptance into the evidence network, a MirrorRecord is validated for: non-empty content_hash; extraction_timestamp within the governed recency window; classification_tags drawn from the schema-defined vocabulary; and absence of a duplicate mirror_id in the local index.

Technical Effect

[0015] Autonomous mirror generation executed within the evidence retrieval pipeline eliminates the manual curation bottleneck, allowing a governed system to expand its verifiable evidence base at the same rate as its retrieval operations without operator intervention.

[0016] Content hash binding at extraction time — before normalisation — ensures that the MirrorRecord captures the exact state of the source at the moment of ingestion, providing a cryptographic reference point for future tamper detection even after the source changes or is removed.

[0017] Governance admissibility gating applied before mirror creation prevents low-quality, failed, or oversized sources from entering the evidence index, maintaining index quality without requiring post-hoc curation.

[0018] Open publication of the MirrorRecord schema and registration procedure as prior art prevents third parties from claiming proprietary rights over the autonomous evidence mirroring surface, ensuring interoperability across AIEP-ecosystem deployments.

Claims

An autonomous evidence mirror generation protocol for a governed AI reasoning system, the protocol comprising: a trigger condition activating mirror generation when evidence retrieval identifies a source URL absent from the local and federated evidence index and the source passes the governance admissibility gate; a content extraction pipeline producing normalised metadata including title, author, publication date, jurisdiction, and classification tags; a content hash computed as SHA-256 over the raw retrieved content body prior to normalisation; and a MirrorRecord comprising a deterministically derived mirror identifier, the source URL, the mirror storage URL, the content hash, the extraction timestamp, and the normalised metadata fields.
The protocol of claim 1, wherein the mirror identifier is derived as the SHA-256 hash of the source URL concatenated with the extraction timestamp, such that the identifier is deterministic given the same inputs and reproducible by any node performing the same mirror operation.
The protocol of claim 1, wherein the MirrorRecord and content snapshot are persisted to a configured durable object store under a deterministic key derived from the content hash prefix and mirror identifier, followed by derivation of an EvidenceIndexEntry and propagation to local and federated index partitions according to the distribution rules defined in the Distributed Evidence Index Protocol.
The protocol of claim 1, wherein MirrorRecord validation rejects records with an empty content hash, an extraction timestamp outside the governed recency window, classification tags outside the schema-defined vocabulary, or a duplicate mirror identifier in the local index, issuing a NegativeProofRecord identifying the rejection reason.
The protocol of claim 1, wherein jurisdiction is inferred from a combination of the source domain’s top-level domain, the detected content language, and explicit jurisdiction metadata present in the source, with inferred jurisdiction values assigned a lower confidence class than explicitly declared jurisdiction values.

Brief Description of the Drawing

FIG. 1 — Mirror generation trigger and pipeline: retrieval → admissibility gate → extraction → hashing → MirrorRecord → persistence → index registration.

FIG. 2 — MirrorRecord schema with field derivation annotations.

Abstract

An autonomous evidence mirror generation protocol triggers structured mirror creation during evidence retrieval when a source URL has no existing MirrorRecord and passes the governance admissibility gate. A metadata normalisation pipeline extracts title, author, publication date, jurisdiction, and classification tags. A content hash is computed over the raw retrieved body prior to normalisation. A MirrorRecord is constructed with a deterministically derived identifier, persisted to durable object storage, and registered as an EvidenceIndexEntry in local and federated index partitions. Validation rejects records with empty hashes, stale timestamps, or invalid classification tags, issuing NegativeProofRecords for failures. The protocol enables AIEP-governed systems to continuously expand their verifiable evidence infrastructure through normal retrieval operations.

Licence

Apache License 2.0 — https://www.apache.org/licenses/LICENSE-2.0