Module 12 · Security, Compliance & Ownership · 12C
Legal, AI & Privacy Controls
Why is a source not collected, and what is the safe fallback?
Try this
- 1Open Maison Ferro KSA decision
- 2Review the fallback and AI controls
- Expected: Shows blocked status, no scheduled run, and the recorded compliant fallback.
Per-source terms/robots decision
Maison Ferro · Saudi Arabia · website
blocked
Decision
Blocked pending legal review (2026-06-03)Scheduled collections
0 — scheduler refuses blocked sourcesrobots.txt
Disallow: /products/Terms
Automated extraction prohibited (§7)Anti-bot policy
Identify user-agent; backoff on 429/503; never solve CAPTCHA, never log in, never rotate identities to evadeSafe fallback
Not yet recorded — choose in Source Due DiligenceSubprocessors & change notices (fictional)
Data handling
Hosting (same-region GCC)
Temporary collection files ≤7 days, same region
OCR vendor (menu boards)
Images only, no personal data; 30-day notice of change
Email relay (digest)
Not used in demo — previews only
AI controls
Matching model governance
- No training on enterprise data — model weights fixed per release
- Human override on every candidate; auto-accept only ≥ threshold and sampled
- Drift sampling: 5% of auto-accepts reviewed weekly (n=10)
- Fairness sampling across brands/markets/languages
- Every decision versioned & attributable
Arabic language quality
Bilingual checks
Arabic title normalisation pass rate 96.2% on 400 sampled listings.
معطر الجسم فاوانيا ← معطر الجسم فاوانيا
Diacritics stripped, alef variants unified, transliteration lexicon (oud/عود, musk/مسك).
Sample PII incident (fictional)
Whitelist & purge
2026-08-02: reviewer comments contained a phone number pattern. PII whitelist allows only product/price fields; 3 comment fields purged; incident closed in 4h.
Whitelist
allowed: [title_en,title_ar,price,currency,size,stock,url_id] blocked: [names,phones,emails,addresses]