Free teaser · No account needed

Will your AI disclosure survive the upload?

Social platforms recompress every upload — decode to pixels, re-encode, and drop every metadata segment: C2PA manifests, XMP packets, EXIF fields. That's label drift: the disclosure label you shipped dies the moment the platform publishes your creative.

Our public drift-test report ran 9 labeled assets through disclosed laboratory simulations of Instagram-, Meta- and TikTok-style pipelines: 9 out of 9 lost their label. This is a lab simulation — not real platform uploads — with the full methodology disclosed. Run the same check on your image below, free, no signup.

Your free drift check

— or —

Simulated Instagram-style recompress (1080px, JPEG Q72). Lab simulation — not a real platform upload.

One free check per visitor · no account · no credit card · file is analyzed in memory and never stored.

Why this matters · Filed 2026-10-12

⏰ The Dec 2, 2026 Article 5 deadline is coming

The EU AI Act's prohibition on non-consensual intimate imagery and CSAM takes effect Dec 2, 2026 — with a top penalty band of up to €35M or 7% of worldwide turnover. Regulators, ad platforms and clients all ask the same question: can you prove your disclosure was in place at publish time? A label that dies on upload is no disclosure at all.

Read the drift-test report →

Sworn statements

Frequently asked questions

What is label drift?

When you upload an image to Instagram, Meta, TikTok or any social platform, the platform recompresses it: decode to pixels, downscale, re-encode as JPEG — and drop every metadata segment (C2PA/JUMBF, XMP, EXIF). If your AI-disclosure lived in that metadata, it dies on upload. Your creative goes live with no disclosure attached.

Is this a real platform upload?

No — and we say so plainly. This check runs your file through a disclosed laboratory simulation of Instagram-style recompression (1080px, JPEG Q72; exact platform qualities are unpublished, so the value is approximated from public reports). It models the same mechanism real platforms use: pixels re-encoded, metadata segments dropped. Full methodology is in the drift-test report.

What do the before/after columns mean?

Before is the provenance fingerprint of your original file — C2PA manifest present/absent, XMP packet present/absent, claim generator, SHA-256 hash. After is the same fingerprint on the simulated published file. Any metadata that vanished is a drift finding.

What do I get if my label died?

Proof that the risk is real, for free. To fix it: serve originals where you control the bytes, keep a publish-time baseline, and re-check the live file — exactly what DisclosureGuard's Drift Monitor automates on Pro: daily re-fetch, baseline diffs, and tamper-evident audit logging.

Why is the Dec 2, 2026 deadline on this page?

The EU AI Act's Article 5 prohibition takes effect Dec 2, 2026, with a top penalty band of up to €35M or 7% of worldwide turnover. Enforcement waves like this are when regulators and platforms start asking for proof — and proof starts with knowing your disclosure survives the publish pipeline.

Is my uploaded file stored or shared?

No. Your file is analyzed in memory for this single check and never stored, never used to train anything, never shared. One free check per visitor, no account, no credit card.