LedgerProof

Methodology

A transparency tool should be transparent. Here is every automated check, its weight, and the exact lexicons the scanner matches. Scoring is the weighted mean of the checks that apply (Pass=1.0, Partial=0.5, Not detected=0); checks marked manual or not-applicable are excluded. Grades: A ≥85, B ≥70, C ≥55, D ≥40, F below 40.

The scanner reads publicly observable signals. Absence of a signal is not a violation, and a present marker is not compliance. A malicious origin can serve different content to the scanner than to users; we do not defend against that and note it here.

CheckWeightWhat it looks for
A50-1-CHAT
Art. 50(1)+(5)
20Conversational UI present + AI-interaction disclosure nearby.
A50-2-C2PA
Art. 50(2)
25Content Credentials (C2PA/JUMBF) coverage across sampled media.
A50-2-PROV
Art. 50(2)
20Valid /.well-known/ai-provenance.json provenance endpoint.
A50-2-META
Art. 50(2)
10Page-level synthetic-content metadata signals.
A50-4-LABEL
Art. 50(4)
10Visible AI-content labeling within the media scope.
A50-5-NOTICE
Art. 50(5)+(1)
15A findable AI-transparency / AI-use notice.

Penalties under the EU AI Act

Fines are tiered by the type of breach — the cap is whichever is higher of a fixed amount or a percentage of total worldwide annual turnover. Article 50 sits in the 3% tier (Art. 99(4)); the Act runs up to €35M / 7% for prohibited practices. For SMEs and start-ups, the applicable fine is the LOWER of the percentage or the fixed amount (Art. 99(6)).

Maximum fineApplies to
€35,000,000 or 7% (Art. 99(3))
Prohibited AI practices
Non-compliance with the Article 5 prohibitions (e.g. social scoring, manipulative or exploitative AI, certain biometric categorisation and real-time remote biometric identification).
€15,000,000 or 3% (Art. 99(4)) Article 50 tier
Operator obligations (incl. Article 50 transparency)
Non-compliance with obligations other than Article 5 by providers, deployers, importers, distributors and notified bodies — including the Article 50 transparency obligations and the high-risk obligations.
€7,500,000 or 1% (Art. 99(5))
Incorrect or misleading information to authorities
Supplying incorrect, incomplete or misleading information to notified bodies or national competent authorities in reply to a request.
€15,000,000 or 3% (Art. 101)
General-purpose AI model providers
Fines the Commission may impose on providers of general-purpose AI models.

Key dates

DateWhat applies
2024-08-01The EU AI Act entered into force.
2025-02-02Article 5 prohibitions and AI-literacy obligations began to apply.
2025-08-02General-purpose AI model rules, governance, and the penalty provisions began to apply.
2026-08-02Article 50 transparency obligations and most high-risk obligations apply. This is the scanner's countdown anchor. this scanner
2026-12-02Transition deadline (per the May 2026 Digital Omnibus provisional agreement) for the Article 50(2) machine-readable marking obligation, for generative systems already on the market before 2 August 2026.
2027-08-02High-risk obligations apply to AI that is itself a regulated product (Annex I) already subject to third-party conformity assessment.

Published lexicons 2026.06-v1

Disclosure phrases: ai assistant, ai-powered, ai powered, automated assistant, virtual agent, virtual assistant, chatbot, automated chat, powered by ai, you are chatting with an ai, this is an automated, ai-generated, ai generated, generated by ai, synthetic media

Chat-widget signatures: widget.intercom.io, intercom, js.driftt.com, drift.com, client.crisp.chat, crisp.chat, code.tidio.co, tidio, js.hs-scripts.com, hubspot, zdassets.com, tawk.to, /chat

Metadata signals: trainedalgorithmicmedia, data-ai-generated, name="ai-generated", compositewithtrainedalgorithmicmedia

Transparency-notice phrases: ai transparency, ai-use, use of ai, use of artificial intelligence, artificial intelligence, automated decision, ai disclosure, responsible ai

Legal/transparency page hints: privacy, legal, transparency, ai-policy, ai policy, terms