AIDEFEND IN ACTION · NOT SCORABLE BY DESIGN

AI Recommendation Poisoning: Summary Links That Try to Rewrite Assistant Memory

This source is retained for threat and defensive-evidence research. Its documented limits do not support a separate deployment score at this record's scope.

Use case sources

Source reports and analysis supporting this case are listed below. AITBM interpretation and any indicative scoring are derived from this public evidence.

Incident overview

Official AIDEFEND in Action September 2026 brief retained as a public-evidence reference; see the linked source for its bounded findings.

Why this analysis has no ERS

Microsoft observed 50 prompt attempts from 31 companies in 60 days, but did not identify an affected assistant/configuration and persistent memory-write outcome for a specific observed attempt. Relecloud is explicitly fictional. Ro-4 and Cn-9 remain unknown.

Status: Not scorable by design · Classification: Incident · source date 2026-09-19

AIDEFEND evidence routes

AIDEFEND defences → AITBM sub-metrics

Identifiers are quoted as they appear on the AIDEFEND in Action brief or in the AIDEFEND SecureFlow defense map (audited AIDEFEND dataVersion 2026.10.15); the sub-metric mapping is AITBM's own, from the specification's AIDEFEND tables reconciled at catalogue data version 2026.10.15. AIDEFEND renumbers identifiers between releases, so the data version travels with every mapping and neither side's IDs should be cited without one. A mapping identifies a possible evidence route; a recommendation does not prove that the control was implemented or effective and receives no scoring credit by itself.

TechniqueDefence PriorityEvidences
AID-D-001.005Recalled Memory Pre-Rehydration ScanningExact upstream page defense ID, verified against AIDEFEND dataVersion 2026.10.15. Presence supplies no score credit.Candidate evidence routeRo-1
AID-E-005Compromised Durable Application Session & Agent State TeardownExact upstream page defense ID, verified against AIDEFEND dataVersion 2026.10.15. Presence supplies no score credit.Candidate evidence routePr-3 Cn-9
AID-H-002.002Inference-Time Prompt & Input ValidationExact upstream page defense ID, verified against AIDEFEND dataVersion 2026.10.15. Presence supplies no score credit.Candidate evidence routeRo-1 Ro-5 Fa-1 Fa-3 Cn-8
AID-I-004.002Persistent Memory Partitioning (Trust & Tenant Isolation)Exact upstream page defense ID, verified against AIDEFEND dataVersion 2026.10.15. Presence supplies no score credit.Candidate evidence routePr-2 Pr-4 Cn-4 Cn-7 Cn-8 Cn-9
AID-I-004.004Transactional Promotion Gates (Quarantine -> Trusted)Exact upstream page defense ID, verified against AIDEFEND dataVersion 2026.10.15. Presence supplies no score credit.Candidate evidence routePr-2 Pr-4 Cn-4 Cn-7 Cn-8 Cn-9

Related AITBM rubrics

Sources