The EU AI Act and your Adsumo creatives

Last updated: August 12, 2026

1. What we mark, automatically

Every image, video and voiceover Adsumo generates carries a C2PA Content Credentials manifest embedded in the file itself. It records that the asset was generated by AI, which model produced it, and when. This happens on every generation, for every customer, in every country. There is no setting to turn it on and nothing you need to do.

Signing happens when the file is stored, before you can download it, and there is no size at which we stop bothering. A 12-second avatar video and a 100MB captioned cut are marked the same way. On the rare occasion a file cannot be signed, we record it as unsigned rather than present it as signed.

This is our obligation, not yours. Article 50(2) requires providers of generative AI systems to ensure outputs are marked in a machine-readable format and detectable as artificially generated. It covers audio, image, video and text alike, which is why our voiceovers are signed and not only the visuals.

2. What we deliberately leave out

The manifest does not contain your prompt, your account, your team, or any identifier that links a creative back to you. It travels on assets you publish to the open internet, so it carries the AI-generated fact and nothing else.

Your prompts routinely describe unreleased products and brand strategy. That is yours, and it stays out of anything we hand you to publish.

3. How to verify it

Upload any Adsumo image or video to contentcredentials.org/verify. You do not need our cooperation to check, and neither does a regulator, a platform, or a researcher.

One honest caveat: we currently sign with our own certificate, so verifiers will show the issuing authority as unrecognised. The content assertions still read correctly and the signature still validates. We are in the process of obtaining a certificate from a C2PA conforming authority, after which the issuer will resolve to Adsumo.

Files you produce with the AI Watermark tool are rendered in your browser, which re-encodes them and strips what was there. We re-sign the result when the source was an Adsumo generation, so those stay verifiable. Above 60MB we store the watermarked file without re-signing it, and files you upload yourself are never given a manifest at all. We do not attach an AI-generated claim to content we did not produce.

4. Whether the Act applies to you

This is the part most people get wrong, so it is worth stating plainly: the Act is not limited to companies based in the EU.

Article 2(1)(c) extends it to providers and deployers "that have their place of establishment or are located in a third country, where the output produced by the AI system is used in the Union." The trigger is where your ads run, not where your company is registered. A US brand running campaigns at German consumers is in scope.

If you advertise to people in the EU, the rest of this page is relevant to you.

5. Your obligation as an advertiser

Article 50(4) requires deployers to disclose AI-generated or manipulated content that constitutes a deep fake. The definition is broader than most people expect:

  • It covers content resembling existing persons, objects, places, entities or events, not only images of people.
  • A realistic depiction of a fictitious person still counts, even with no real individual involved.
  • The test is whether an average person would be misled about authenticity. In practice that means photorealism. Clearly unrealistic content, such as fantasy scenes, falls outside.
  • Advertising does not qualify for the artistic or creative exemption. "It's an ad" is not a defence.

We provide a Watermark tool that adds a visible AI-generated label to images and video, and it preserves the Content Credentials rather than stripping them. You can label one creative or a batch of up to 25 in a single run, and the same tool is available to your AI agent over MCP, so a campaign assembled by an agent can be watermarked without a human opening the dashboard.

The tool can apply either your own wording or the European Commission's official AI-labelling icons, which distinguish content made entirely by AI from your own content that AI modified. The Commission notes that using the icons does not by itself establish compliance.

We do not apply it for you, and we do not decide which of your creatives need it. Whether a given ad requires disclosure depends on how you publish it, the claims it makes and the audience it reaches. We cannot determine that from the pixels, and telling you a creative is fine when it is not would be worse than saying nothing. This page is information, not legal advice.

6. Timing

The Article 50 transparency obligations apply from 2 August 2026. Under the AI Omnibus, generative systems already on the market before that date have until 2 December 2026 to meet the machine-readable marking requirement.

7. Read it yourself

This page is our summary, and a summary is a thing with a point of view. The law is public and short enough to check us on.

8. Questions

Email help@adsumo.ai. If you need a written statement of our marking practices for a compliance review or a platform audit, ask and we will provide one.