Verifiable Forgetting in Multi-Tier Memory Systems via Tier-Specific Mutual-Information Thresholds — A Technical Note

Defensive Publication establishing prior art under § 3 Abs. 1 PatG, Art. 54 (2) EPC, and 35 USC § 102 (a)(1) for a tier-specific mutual-information threshold method for verifying record deletion in multi-tier memory systems used in agentic AI products. The method addresses the regulatory requirement of demonstrable data deletion under GDPR Article 17 and the EU AI Act Article 10 in scenarios where conventional procedural deletion logs are insufficient because residual statistical information about deleted data may persist in derived representations such as embeddings, attention weights, or fine-tuned parameters. The technical note describes a tier-specific mutual-information threshold chain combining a configurable estimator family (KSG, MINE, InfoNCE, BMA-MI, CMI-NN) with a sensitivity-classification-driven threshold lookup and an automatic re-training, re-indexing, fallback-to-purge, or regulatory escalation trigger. Seven explicit variants and eight claim-style disclosure statements are included. Released under CC0 1.0 Universal Public Domain Dedication. No patent rights are claimed by the author or any affiliated party regarding the disclosed embodiment. HEINI® is a registered trademark of Daniel Hans-Georg Heinen at the German Patent and Trademark Office (DPMA), registration number 30 2026 214 789, classes 09, 35, 42, registered 25.03.2026. Canonical URL: https://heini.app/en/blog/verifiable-forgetting-multi-tier-memory