AI Governance Policies
Cardinal #112 tangible-build · Pieces 100-105 + 110 consolidated · 7 policies active
What this vessel does
AI tool usage across KolBo customer devices is governed by 7 policy pieces. This vessel surfaces all 7 + per-tier defaults at basic functional level. Piece 100 (Shadow AI Detection) is now mutation-backed — toggle writes to customer_rulebook.shadow_ai_detection + emits an audit event. Pieces 101-105 + 110 remain static descriptors pending their own mutation reps.
Shadow AI Detection
Piece 100 · mutation-backed
Detect ChatGPT/Claude/Gemini/Perplexity usage across family-member devices. Domain + app + API-endpoint detection layers.
PII Detection in AI Prompts
Piece 101
Pre-submission scanning for phone/SSN/address/email/frum-community PII (shul/yeshiva/rebbe/family names).
AI Prompt Content Scanning
Piece 102
Harmonic-pattern frum-context categories: inappropriate-content + lashon hara + halachic-question + anti-frum + time-wasting.
AI Tool Allowlist Per Tier
Piece 103
Sefaria AI + Aleph Beta + ParshaBot universal. Business AI per tier. Character.ai blocked all tiers.
AI Response Content Filtering
Piece 104
Browser Extension DOM hooks + classifier (Piece 76 composition). KolBo Approved rabbinic resource positive-redirect.
Prompt Injection Defense
Piece 105
URL-content classifier + injection-pattern detection + sandbox AI responses + context isolation.
AI Tool Category Taxonomy
Piece 110
12 KolBo AI categories: General + Torah + Writing + Image-gen + Code + Video-gen + Voice + Relational + Search + Workflow + Image-edit + Music. Per-tier defaults matrix.
Phase 2 chazarah: per-policy operator toggle + per-tier customization UI + per-customer override path + AI Governance audit log surface. Lane B3 ships Piece 100 mutation handler; depends on Lane B2 schema 0004 migration applying for runtime correctness against live Supabase.