Tested BMW, Stadt Koeln, BfDI, Sparkasse, Caritas, TUEV Sued,
Spiegel, ETO Gruppe, EUIPO. Key findings:
- Stadt Koeln + ETO Gruppe best (95% correctness)
- BMW, Sparkasse, Spiegel genuinely deficient (verified)
- EUIPO uses EU Regulation 2018/1725, not GDPR — needs separate checklist
- ~0-2 false positives per website after LLM verification
7 regex fixes emerged from batch testing (soft hyphens, word
insertions, numbered headings, German section names, etc.)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Detailed plan for upgrading the 22 existing Playwright-based banner
checks to the same quality level as the document checks:
- 6 L1 + 30 L2 hierarchical checks
- Expert hints with EuGH/CNIL/DSK/EDPB references
- 3-phase evidence (before consent, after reject, after accept)
- Dark pattern detection (button size, color, click asymmetry)
- Estimated 3-4h implementation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fixes based on manual verification of all 30 failed checks:
1. Cookie table: recognize "folgende cookies" + column headers as text
2. Cookie names: add JSESSIONID, cookieinfo, et_id, BT_* patterns
3. Essential justified: match "sitzung zuordnen", "betrieb der website"
4. Social bookmarks: recognize as 2-click alternative
5. DSFA plural: "kanaelen" now matches alongside "kanal"
6. Section splitter: skip-headings no longer lose subsequent text
(Risikoabwaegung section was cut from DSFA, losing risk scores)
7. Cookie legal basis: accept Art. 6(1)(f) in cookie context
Reduces false positives from 7 to ~1-2 for IHK Konstanz test case.
Ground truth table: zeroclaw/docs/ground-truth-ihk-konstanz.md
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Only Cookie and Widerruf sections are checked as separate documents.
Social Media, DSFA, Betroffenenrechte, Dienste von Drittanbietern are
part of the parent DSI and no longer generate false findings.
Added PLAN-rag-document-check.md for Phase 2:
- RAG-based checks with document-type-specific Controls
- DSFA checklist (Art. 35 + Landes-Listen)
- AVV checklist (Art. 28)
- Reference detection (sub-doc → parent doc)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Automated comparison: services mentioned in privacy policy vs. actually
embedded on website. Three categories: undocumented (Art. 13 violation),
outdated (cleanup), correctly documented (check third country only).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Multi-page crawl: scan 5-10 strategic pages (start, footer links) for
chatbot widgets, AI text mentions, and tracking services. Feed results
into relevance filter to reduce false positives.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses false-positive controls like C_TRANSPARENCY being recommended
when no AI usage is evident. Plan for separate implementation session.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Benjamin Admin