Policy · 4 June 2026 · 8 min
AI and the EU AI Act: how we label commercial imagery
From 2026 the EU AI Act requires clear recognition of AI-generated imagery. This is how we handle that in commercial production, practically.
The EU AI Act was adopted in 2024 and the first rules on synthetic imagery are being strictly enforced from 2026. For brands working with AI product photography that raises legitimate questions: does every image need a sticker, can you still use recognisable models, and who's responsible?
What the law does and doesn't ask
The AI Act requires providers of AI systems to make synthetic imagery machine-readable as such. That's an obligation on the model makers, not on you as a brand. What is your responsibility: not misleading consumers.
A product photo of your actual packaging, in a generated setting, isn't misleading as long as the product is accurate. A 'before and after' image in skincare sits more sensitively and asks for clear disclosure when it's AI-generated.
How we handle it in production
Every series we deliver carries C2PA metadata inside the image files. That's an open standard showing the image had AI-involved production, including timestamp and model used. Invisible to the viewer, verifiable for anyone checking.
On top of that we hold a few internal rules: no realistic faces of existing people without consent, no comparable results promised that the product can't deliver, and for sensitive contexts (medical, children, food claims) a manual check before publication.
What this means for you as a client
Practically you barely need to do anything different from classic imagery. Disclose AI involvement when it's relevant to the consumer (typically with before-after, demonstrations of effect), and keep the metadata we deliver inside your asset system.
Our approach page shows how we've embedded these checks into the production flow. In our packages C2PA labelling is included by default, at no extra cost.
Keep reading