Public evidence directly supports the EU AI Act Article 50 observation.
Methodology
How isGenAI measures EU AI Act Article 50 transparency.
isGenAI is an external public-evidence rating system. It records what can be independently observed, what requires context, and what changed over time.
V1Methodology version
2026Scanner registry
5EU AI Act Article 50 control families
5Standardized checks
Status taxonomy
Observed status is separate from legal compliance.
isGenAI avoids collapsing ambiguous public evidence into binary claims.
Some public signal exists, but it is incomplete, narrow or implementation-limited.
The scan did not observe a public signal. This is not a legal non-compliance finding.
External testing cannot verify the control from public evidence.
Vendor-submitted evidence has passed isGenAI review and is marked as vendor-originated.
A correction or dispute is open and the affected observation is flagged.
The control is not applicable to the assessed product context.
Legal or deployment context is required before an external inference is fair.
Scoring
Score, grade and confidence
VERIFIED and VENDOR VERIFIED count as 1. PARTIAL counts as 0.5. NOT OBSERVED counts as 0. UNVERIFIABLE, REQUIRES CONTEXT, DISPUTED and NOT APPLICABLE are excluded from the denominator.
A90-100
B80-89
C70-79
D60-69
E50-59
F0-49
Confidence is calculated separately from the rating using direct testability, evidence completeness, scanner confidence, freshness and excluded-control share.
EU AI Act Article 50 model
Controls included
EU AI Act Article 50(5) is treated as a cross-cutting clarity, timing, distinguishability and accessibility requirement, not a separate marking obligation.
Rule registry
Versioned source registry
Every assessment stores raw observations and the derived rating with methodology and scanner versions.
Regulation (EU) 2024/1689 - Artificial Intelligence ActEuropean UnionEffective 2024-08-01 / status CURRENTGuidelines on transparency obligations for providers and deployers of AI systemsEuropean CommissionEffective 2026-08-02 / status CURRENTCode of Practice on Transparency of AI-generated ContentEuropean Commission / AI OfficeEffective 2026-08-02 / status CURRENTWeb Content Accessibility Guidelines 2.2W3CEffective 2023-10-05 / status CURRENT
isGenAI records externally observable technical evidence and separates legal applicability from public observations.
EXTERNALLY_TESTABLE / reviewed 2026-08-16Machine-readable marking is modality-specific; isGenAI records observed mechanisms and tested persistence, not universal robustness claims.
EXTERNALLY_TESTABLE / reviewed 2026-08-16isGenAI records externally observable technical evidence and separates legal applicability from public observations.
CONTEXT_REQUIRED / reviewed 2026-08-16isGenAI records externally observable technical evidence and separates legal applicability from public observations.
EXTERNALLY_TESTABLE / reviewed 2026-08-16isGenAI records externally observable technical evidence and separates legal applicability from public observations.
PARTIALLY_TESTABLE / reviewed 2026-08-16Quality gate
Only evidence-rich scorecards are indexed.
isGenAI requires a canonical organization, product classification, official domain, standardized scan, evidence records, methodology version and publication review before a public Scorecard is indexable.
A failed fetch, access challenge or empty page extract cannot support a public grade. Scored output-marking checks require an artifact inspection; public marketing language alone cannot establish that marking is present or absent. Verified accessibility and timing require evidence from an accessibility check or a reviewed workflow. Assessments that lack this evidence are retained for review, while their ratings and related public findings are withheld.