The label is gone, the obligation remains
One screenshot is enough and the AI image keeps circulating unlabeled. With on:mint, you can prove at any time that you labeled your asset correctly.
Label AI content. Get the obligation right.
As of August 2, 2026, the EU AI Act requires AI content to be labeled. Visible to people, readable by machines. mintys, the labeling tool by on:mint, does both in one step.
Not sure whether this applies to you?Take the self-test
Legal tech tool for technical AI labeling
Based on the C2PA metadata standard
Data and signature infrastructure in the EU
Labeling is mandatory.
Proving is better.
Since August 2026, AI content must be labeled: in machine-readable form by providers, visibly by deployers. But a visible label can be removed, overwritten or cropped out, and many lack the infrastructure for technical marking. Either way, the question remains: how do you prove that you labeled correctly?
Machine-readable is easier said than built
The AI Act requires marking that is effective, interoperable, robust, and reliable. Building that yourself takes months – and the standard it will be tested against is only just emerging.
The chain breaks at the first download
AI provider, agency, brand, marketplace: every station must meet its own obligations. Without provenance information traveling with the asset, no one can reliably tell what is AI and what was changed.
For labeling, there's mintys.
mintys is our labeling tool: upload the image, add the label, keep posting. The visible AI label and the machine-readable marking are created in one step and stay with the image, wherever it ends up. Prices and all the details are on the mintys site.
The labeling obligation is already in force.
In force since August 2
Anyone publishing AI content that could pass for real has had to disclose it visibly since August 2, 2026. Providers of generative AI systems have until December 2, 2026 to put machine-readable marking in place. Violations can be fined up to €15 million or 3% of worldwide annual turnover.
Platforms are moving first
LinkedIn and other platforms already display Content Credentials; the EU is working on a standardized AI label. Sign now and be compatible before everyone else is.
Your assets are already circulating
AI images sit in shops, campaigns and on marketplaces, mostly without provable provenance. What goes out unsigned today can't be proven after the fact.
Make
your AI labeling
provable
Labeling has been mandatory since August 2, 2026. With mintys, you get it done in one step and can prove it at any time.
Would you rather check what applies to you first?Take the self-test
FAQs
Since August 2, 2026. Providers of generative AI systems must mark outputs in a machine-readable format, recognizable as artificially generated or manipulated (Art. 50(2)). Deployers must disclose deepfakes in a perceivable way, as well as AI-generated text on matters of public interest published without editorial responsibility (Art. 50(4)) – at first exposure at the latest, clearly recognizable, not in the fine print. For the machine-readable marking under Art. 50(2), providers of generative AI systems have until December 2, 2026 to put the required measures in place.
Providers develop an AI system or place it on the market under their own name; deployers use it professionally under their own responsibility. Important: even companies that merely use AI tools – say, a brand or agency creating and publishing AI images – count as deployers and carry the obligation. As a provider, you implement machine-readable marking with on:mint via API or MCP; as a deployer, you secure your visible disclosure cryptographically and steganographically and can prove it at any time.
It meets the disclosure obligation — but it is only a claim: removable, overwritable, undocumented. If the asset keeps circulating without the label, you lack proof that you labeled it correctly. Content Credentials make labeling tamper-evident and provable.
Content Credentials are the tamper-proof passport for digital assets: a signed, machine-readable provenance record documenting who created an asset, with which tool, and what was changed. The foundation is the open C2PA standard backed by Adobe, Microsoft, and LinkedIn, among others; on:mint extends the standard with a cryptographic hash, watermark, perceptual fingerprint, deepfake detection, and an immutable anchor, so the label stays provable even after the file has been edited or re-compressed. That restores transparency and gives digital assets a clear quality signal again. It becomes visible via the Cr icon: one click shows the full provenance history.
We convert them: existing metadata is transferred into the C2PA format, and entire holdings can be signed and verified in batch.
Three ways: the web app for single files with no developer effort, the API for batch processing and production pipelines with usage metering, and MCP for integration into agent workflows.
Our verdict is honest: if content isn't verifiably AI, we don't label it as AI. That also protects your genuine originals from false suspicion.
The infrastructure is located in the EU. Anyone can read the credentials – that is the point of an open standard. You control which details become public: sensitive metadata such as exact GPS locations does not have to be published. Only authorized parties can create and sign credentials; on:mint signs as a C2PA issuer.
Questions? We're happy
to answer them.
