EU AI Act · Article 50
LedgerProof turns every AI-generated output into a receipt anyone can verify independently — cryptographically anchored, checkable in seconds, without trusting LedgerProof or any vendor. Article 50 asks you to prove what your AI did. A log you wrote yourself can't. This can.
Article 50 transparency obligations apply · — days · 2 Aug 2026
Article 50(1)
Receipts capture every AI-mediated user interaction with the disclosure language used, signed by the deployer and cryptographically anchored. The receipt format rejects email addresses in policy-protected fields at parse time — hash-only and GDPR-safe by construction.
Article 50(2)
Receipts cover image, audio, and video outputs with a content hash, the generation context, and the marking mechanism applied — designed to interoperate with C2PA and OpenTimestamps for defense-in-depth across the synthetic-media ecosystem.
Article 50(4)
Receipts attach to AI-generated text and deepfake artifacts your organization publishes or transmits — chatbot responses, generated copy, summaries, automated decisions. The SDK wraps any model call with negligible overhead.
Article 50 enforcement begins 2 August 2026, and the EU Code of Practice is still settling. Most organizations don't have a defensible, independent audit trail for the AI deployed inside their own perimeter. The honest answer to "what is our exposure?" is usually: "we don't fully know" — because the only evidence is a log the organization wrote itself, which proves nothing to a regulator.
LedgerProof is not a vendor solution that asks you to trust it. It's an open protocol that implements the IETF SCITT architecture (Internet-Draft draft-dawkins-scitt-ai-article50), stewarded by a California nonprofit (501(c)(3) application in process). The receipt format is open, the verifier is open-source, and you can self-host the operator if procurement requires it. If LedgerProof Inc. ever stops, the protocol — and every receipt you hold — keeps verifying against the public chain.
For the moving Code of Practice: the receipt schema uses a profile system. When the Code specifies a new disclosure obligation, a new profile is published. Existing receipts stay valid under their original profile; new outputs flow into the updated one. You don't re-paper your AI deployment every time the regulator moves.
Free · 60 seconds
Run a free Article 50 scan against any public AI surface you own. It grades you against the obligations and shows the gaps — no account, nothing to install.
Run a free scanThe system of record
Register your AI systems, monitor them continuously, anchor the evidence, and export an auditor-ready dossier. Priced per AI system — from a self-serve plan to enterprise.
See pricingWhen it's contested
When a single proof has to perform — a regulator's request, a litigation exhibit, an audit workpaper — the Verified Evidence Package turns it into a dispute-grade bundle a third party can re-verify.
The Verified Evidence PackageLedgerProof produces independently verifiable evidence, not a verdict — your auditor or counsel determines compliance and admissibility. Proofs are tamper-evident, not tamper-proof. Not legal advice.
The open verifier is live. The protocol is published as an IETF Internet-Draft and implements the SCITT architecture. Article 50 transparency obligations apply from 2 August 2026.