The hero image came out of Midjourney, the explainer used a synthetic presenter, half the product copy was drafted by a model, and the voice on the podcast read was never in the room. Six months later somebody asks which of it carries machine-readable marking, which needed a visible disclosure, and which deadline applied. Nobody kept a list. This is the list — one row per asset, with the verdict, the deadline and the fix already worked out.
Every asset gets a verdict, not just a row. The dashboard shows how much of your in-scope output is actually marked, which deadline each gap falls under, and what to fix first.
A spreadsheet records what you typed. This works out the verdict for you, applies the right deadline to each asset, and gives you something dated to hand over.
Article 50 asks two different things: machine-readable marking on the output, and a clear visible disclosure where the content shows realistic people, voices, places or events. An asset can pass one and fail the other, so the register scores them apart and only demands a visible disclosure where it applies.
Obligations bite immediately for systems placed on the market after 2 August 2026, and on 2 December 2026 for tools already on the market before that date. Grandfathering follows the generating system, not the asset, so each entry records which basis applies and counts down against it.
Every row names the model or tool that produced the asset. When a vendor turns marking on, or you find one that never emitted it, you can filter straight to every affected asset instead of guessing which campaigns used which tool.
CSV for the compliance file or a shared drive, JSON for a backup you can reload. The export carries the verdict and the reasoning beside each asset, so the person reading it six months later does not have to reconstruct your thinking.
Not for counsel, and not for a governance platform nobody in marketing has a login to.
You commissioned the AI creative and you own the answer when someone asks what was generated and whether it was labelled. Keep one register per brand and update it as assets ship.
Log each client's assets separately and hand over a dated CSV with the campaign. A marking record attached to delivery is cheap to produce and awkward to be without.
You are the one who finds out a tool never emitted provenance metadata. Filter by tool, see every affected asset, and work the remediation list rather than a memory of which campaigns used it.
Whether it is a client audit, an internal sign-off or a regulator's question, the useful artefact is a dated register showing what you checked and when. That is the whole output.
“We know roughly half of it was AI. We could not tell you which half.”
Enterprise AI-governance platforms are priced for enterprises and sold to a risk team. Counsel bills by the hour and starts by asking you to list the assets — the list you do not have. Between those two there was nothing structured, cheap and owned by the marketing team itself — that is what this is.
One self-contained HTML file. No accounts, no API keys, no network calls, nothing uploaded. It installs to your phone or desktop home screen, works offline, and every entry stays in your own browser. Built by Mulkern AI Systems, who ship AI operations tooling for marketing teams.
A structured record-keeping aid, not legal advice. It helps you find the gaps and decide what to take to counsel — it does not replace them.
Instant access after checkout. No refunds. All sales are final.