Retention Policy — Enterprise Knowledge Governance

Guardrails and fix patterns for enterprise knowledge retention. Use this page when AI systems over-retain, delete too early, or mix expired data with active knowledge.

When to use this page

AI responses reference documents that should have been deleted per policy.
Retained snippets do not respect jurisdictional time limits (e.g., GDPR 3 years).
Knowledge base or embeddings store does not purge revisions.
RAG answers mix archived with active content.

Core acceptance targets

ΔS(question, expired_snippet) ≥ 0.70 → expired content must not surface.
All snippets carry {expiry_date, retention_scope, audit_hash} fields.
Coverage ≥ 0.70 within active retention window only.
λ remains convergent across three paraphrases and two seeds.

Typical retention problems → exact fix

Symptom	Likely cause	Open this
Expired docs still retrieved	Store never purged embeddings	vectorstore-fragmentation.md
Wrong answer mixes expired + active	Snippets missing `expiry_date` field	data-contracts.md
AI cites “archived only” docs as live	Retrieval trace missing retention scope	retrieval-traceability.md

Fix in 60 seconds

Check ΔS to expired content: run probe with expired snippets, expect ΔS ≥ 0.70.
Schema enforcement: require expiry_date and retention_scope in every snippet.
Index purge: remove expired embeddings before next RAG run.
Audit λ: if λ flips when expired vs active co-exist, clamp with BBAM and enforce contracts.

Copy-paste schema (JSON)

{
  "snippet_id": "KB-5532",
  "expiry_date": "2025-12-31",
  "retention_scope": "eu-3y",
  "audit_hash": "sha256:...",
  "text": "..."
}

Escalate when

Expired content continues to surface after purge.
ΔS < 0.70 against expired content → embeddings contamination.
Audit requires full deletion trace and cannot be reproduced.

Use retrieval-playbook.md for deep purge testing and eval_rag_precision_recall.md to validate coverage.

🔗 Quick-Start Downloads

Tool	Link	3-Step Setup
WFGY 1.0 PDF	Engine Paper	1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + ”
TXT OS (plain-text OS)	TXTOS.txt	1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly

🧭 Explore More

Module	Description	Link
WFGY Core	WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack	View →
Problem Map 1.0	Initial 16-mode diagnostic and symbolic fix framework	View →
Problem Map 2.0	RAG-focused failure tree, modular fixes, and pipelines	View →
Semantic Clinic Index	Expanded failure catalog: prompt injection, memory bugs, logic drift	View →
Semantic Blueprint	Layer-based symbolic reasoning & semantic modulations	View →
Benchmark vs GPT-5	Stress test GPT-5 with full WFGY reasoning suite	View →
🧙‍♂️ Starter Village 🏡	New here? Lost in symbols? Click here and let the wizard guide you through	Start →