◎ OS PUB Apache 2.0 ← All specifications

P147 — AIEP — Evidence Freshness Decay and Expiration Protocol

Publication Date: 2026-03-27 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, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure defines a time-dependent freshness model for AIEP evidence artefacts, specifying source-class-specific freshness thresholds, a governed expiry transition sequence, and a re-acquisition pipeline that replaces expired artefacts with verified current versions without breaking existing reasoning chain citations.

Field of the Disclosure

[0003] This disclosure relates to evidence freshness lifecycle management protocols for governed artificial intelligence reasoning systems.

[0004] More particularly, the disclosure concerns a freshness decay model under which each evidence artefact carries a freshness_score that declines from 1.0 at admission toward 0.0 as time elapses, with governed transitions at defined thresholds from FRESH to STALE to EXPIRED, and a re-acquisition workflow triggered before expiry.

Background

[0005] Evidence artefacts admitted to the AIEP evidence rail at a point in time may become unreliable as the underlying information changes. A regulatory document from two years prior may have been superseded; a news article’s facts may have been corrected; a dataset may have been updated with revised figures. Without a freshness model, an AIEP-governed system cannot distinguish between a newly admitted artefact and one admitted three years ago and never rechecked.

[0006] Existing approaches treat evidence as either permanently valid until explicitly retracted, or require manual staleness review. Neither approach is adequate for high-frequency evidence bases: perpetual validity risks reasoning on outdated information, while manual review does not scale across thousands of artefacts.

[0007] No existing AIEP specification defines the parameters of a freshness decay model, the source-class-specific thresholds at which artefacts transition between freshness states, or the governed re-acquisition pipeline that maintains evidence continuity across the transition.

Summary of the Disclosure

[0008] Each EvidenceNode carries freshness_score (float in [0.0, 1.0]), freshness_state (enumeration: FRESH, STALE, PRE_EXPIRY, EXPIRED), admission_timestamp, and freshness_class (source-class-specific decay parameters). Freshness score is computed as:

freshness_score = max(0.0, 1.0 − (elapsed_seconds / class_ttl_seconds))

where class_ttl_seconds is the configured time-to-live for the artefact’s source class.

[0009] Source Class Freshness Parameters define class_ttl_seconds and threshold values per source class:

Source Class	TTL	STALE threshold	PRE_EXPIRY threshold
Regulatory	365 days	0.30	0.10
News	7 days	0.30	0.10
Academic	730 days	0.25	0.05
Dataset	90 days	0.30	0.10
Sensor	1 day	0.50	0.10
General Web	30 days	0.30	0.10

All thresholds are schema-defined and may be overridden per-node through the governance configuration.

[0010] State Transitions: A FreshnessTransitionRecord is produced by a scheduled FreshnessMonitorWorker whenever an artefact transitions from FRESH to STALE, STALE to PRE_EXPIRY, or PRE_EXPIRY to EXPIRED. Transitions are deterministic: the transition timestamp is computed as admission_timestamp + (threshold × class_ttl_seconds).

[0011] On transition to PRE_EXPIRY, the FreshnessMonitorWorker queues a re-acquisition task: fetch the source URL, compute the current content hash, and compare against the admitted content_hash. If hashes match, the artefact is renewed: freshness_score resets to 1.0, admission_timestamp updates, and a RefreshRecord is appended. If hashes do not match, the Continuous Evidence Background Learning Protocol (P145) drift pipeline is invoked.

[0012] On transition to EXPIRED, the artefact’s freshness_state is set to EXPIRED and the artefact is flagged as inadmissible for new reasoning chains. Existing reasoning chains that cite the artefact retain their citation but receive an EXPIRED_EVIDENCE_WARNING when the chain is next evaluated. The artefact is not deleted — it remains archived and accessible via the global mirror network (P142) for audit purposes.

[0013] A FreshnessPolicy governs whether reasoning chains may cite STALE artefacts. The default policy permits STALE citations with a stale_evidence_warning. An operator may configure the policy to REJECT_STALE, in which case any reasoning chain that attempts to cite a STALE or EXPIRED artefact is blocked with a governance rejection record.

Technical Effect

[0014] The deterministic freshness computation from admission timestamp and TTL — rather than event-driven staleness flags — ensures all nodes independently compute the same freshness state for a given artefact, without requiring synchronisation.

[0015] Per-source-class TTL configuration reflects the empirically different rates at which different evidence categories become unreliable: sensor data ages in hours; academic citations remain valid for years.

[0016] Archiving rather than deleting expired artefacts preserves audit trail integrity: reasoning chains produced during the artefact’s FRESH period remain fully auditable without requiring the expired artefact to remain admissible for new chains.

Claims

An evidence freshness decay and expiration protocol for a governed AI reasoning system, the protocol comprising: a freshness score computed as a linear function of elapsed time relative to a source-class-specific time-to-live parameter; state transitions from FRESH to STALE to PRE_EXPIRY to EXPIRED at schema-defined freshness score thresholds; and a FreshnessTransitionRecord produced by a scheduled monitor worker at each state transition.
The protocol of claim 1, wherein a PRE_EXPIRY transition triggers a re-acquisition task that fetches the source URL, computes the current content hash, and renews the artefact with a reset freshness score if hashes match, or invokes the drift detection pipeline if hashes differ.
The protocol of claim 1, wherein EXPIRED artefacts are archived rather than deleted, preserving audit trail integrity for reasoning chains produced during the artefact’s FRESH period, and wherein new reasoning chains are blocked from citing EXPIRED artefacts.
The protocol of claim 1, wherein a FreshnessPolicy governs whether reasoning chains may cite STALE artefacts, with a configurable REJECT_STALE mode blocking any new citation of a STALE or EXPIRED artefact with a governance rejection record.

Brief Description of the Drawing

FIG. 1 — Freshness state transition diagram: FRESH → STALE → PRE_EXPIRY → EXPIRED with threshold values and re-acquisition branch.

FIG. 2 — Freshness score decay curve over time for three source classes (Regulatory, News, Sensor) showing different TTL slopes.

Abstract

An evidence freshness decay and expiration protocol assigns each AIEP evidence artefact a freshness score computed as a linear function of elapsed time relative to a source-class-specific time-to-live. A scheduled FreshnessMonitorWorker produces FreshnessTransitionRecords as artefacts move from FRESH to STALE to PRE_EXPIRY to EXPIRED. On PRE_EXPIRY, a re-acquisition task verifies the source and renews the artefact if unchanged, or invokes the drift detection pipeline if changed. EXPIRED artefacts are archived rather than deleted to preserve reasoning chain audit trails. A FreshnessPolicy governs whether STALE artefacts may be cited in new reasoning chains.

Licence

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