EU AI Act · Article 50 · updated for 2026

AI photo editing and EU rules: what changes when an image is generated or manipulated.

The EU AI Act transparency rules under Article 50 apply from 2 August 2026. For website owners, photographers and designers, the important distinction is between ordinary assistive editing and AI-generated or manipulated content that can mislead people about what is authentic.

Usually lower concern

Assistive standard editing

Examples can include normal cleanup, exposure, colour, noise reduction and other edits that do not substantially change the input or its meaning.

Needs closer review

Substantial synthetic changes

Generated people, places, events or materially altered scenes can move into Article 50 marking or disclosure territory.

The short version

Not every AI-assisted edit needs a visible “AI” label.

Article 50 distinguishes between provider-side marking of synthetic content and deployer-side disclosure of certain uses, including deepfakes. The Commission’s 2026 guidance also recognises an exception where AI performs an assistive function for standard editing or does not substantially alter the user’s input or its semantics.

Practical website rule

If an edit simply helps a real photograph look cleaner or more polished without changing what the scene actually represents, that is very different from generating a new person, replacing a real place with a fictional one, or creating a realistic event that never happened.

Examples for website photography and design.

Exposure, colour, crop and cleanup

Normal photographic adjustments are generally closer to standard assistive editing. The AI Act’s provider marking obligation expressly contains an exception for assistive standard editing or changes that do not substantially alter the input data or its semantics.

Removing a small distraction

Removing dust, a temporary object, skin blemishes or a minor background distraction may still be ordinary retouching depending on the effect. The key question is whether the edit materially changes what a viewer would believe about the person, product, place or event.

Changing the background colour

A simple colour change may remain a design edit if it does not create a false factual impression. But replacing the entire real environment with a realistic invented office, destination or event is a much more substantial manipulation and should be reviewed differently.

Generating a new realistic background

This can become more sensitive when the result resembles a real place, entity or event and would falsely appear authentic. Under Article 50, deployers must disclose AI-generated or manipulated image, audio or video content that constitutes a deepfake.

Generating or materially changing a person

If AI makes a real person appear to do, wear, attend or experience something that did not happen, the risk of the content being treated as a deepfake is much higher. For commercial website use, preserving a clear line between real documentation and synthetic creative imagery is usually the safer design choice.

What this means for Create23 website projects.

For original website photography, the default approach is to preserve the reality of the person, workplace, product or location being documented. AI can be useful for restrained editing, but it should not quietly turn authentic business photography into a fictional representation of the business.

Where a project intentionally uses synthetic or materially manipulated imagery, the image should be assessed separately for transparency requirements, context and whether a viewer could reasonably mistake the result for authentic content.

This page is a practical guide, not legal advice.

The exact application of the AI Act depends on the tool, the role of the provider or deployer, the kind of manipulation and how the content is published. For high-risk or ambiguous uses, use the official Commission guidance or obtain legal advice.

Official sources.

AI photo editing EUEU AI Act photographyAI image disclosure EUAI edited photos website