The EU AI Icon Is Coming: Soon a Label on All Content Generated by Artificial Intelligence

Soon, a badge on all your AI-generated images, videos and text? That's the direction Europe is heading. On 5 March 2026, the European Commission published the second draft of the Code of Practice on the marking and labelling of AI content — and it goes much further than the first. The distinction between "fully AI-generated" and "AI-assisted" content? Gone. The objective is now clear: one icon, one treatment, for all content touched by AI.

What's changing in the 2nd draft

The first draft Code of Practice, published in December 2025, proposed a two-tier taxonomy: "fully AI-generated" for content created without human involvement, and "AI-assisted" for content involving significant human contribution. This distinction immediately raised thorny legal and commercial questions — particularly around copyright: content labelled "fully AI-generated" risked losing its protection.

The second draft, prepared by independent experts after hundreds of contributions from industry, academia and civil society, settles the matter: the taxonomy has been completely removed. There is now only a single labelling regime, applied uniformly to all content falling under Article 50 of Regulation (EU) 2024/1689.

In practice, the 3 major changes:

• No more "fully AI" vs "AI-assisted" distinction — A visual retouched with AI and an image generated from scratch follow the same labelling rules
• Open standards promoted — The code encourages the use of open standards for technical marking, reducing costs and fostering interoperability
• Reduced compliance burden — Consolidated measures, optional elements, and more flexibility to adapt solutions to each signatory's context

The European AI icon: first visuals

The annex of the second draft includes for the first time illustrative examples of a potential European icon, to be made freely available to signatories. The code also proposes the creation of a task force to develop a harmonised interactive icon at EU level.

Pending the definitive version, the chosen solution is a symbol containing the acronym "AI" (or its local equivalent: "IA" in French and Spanish, "KI" in German). The icon must be:

• Visible from the first exposure to the content
• Positioned consistently — in a fixed location appropriate to the context
• Compliant with accessibility requirements under EU law

The precise details (size, colour, placement) will be discussed with stakeholders in upcoming workshops. But the direction is clear: ultimately, all AI content published in Europe will be accompanied by an identifiable visual marker.

Why removing the taxonomy is controversial

The disappearance of the "fully AI" / "AI-assisted" distinction simplifies compliance, but it creates new tensions.

For creators: the risk of stigmatisation

A photographer who retouches a landscape with a generative fill tool and an account publishing entirely synthetic images will now carry the same label. For creators who use AI as one tool among many, this uniformity can feel like a stigmatisation of any AI use, without nuance.

For businesses: a welcome simplification

The two-tier taxonomy posed considerable practical problems: how do you determine the threshold of human contribution? Who decides between "fully AI" and "assisted"? The removal eliminates these grey areas and reduces the risk of misclassification.

For copyright: a relief

The most sensitive question from the first draft — does content labelled "fully AI-generated" lose its copyright protection? — disappears with the taxonomy. The label no longer says how much AI was used, only that AI was used. The question of originality and human contribution remains settled by copyright law, not by labelling.

What's planned for artistic works

The second draft refines the regime applicable to evidently artistic, creative, satirical or fictional works. These works are not exempt from labelling, but benefit from lighter rules that facilitate integration with existing practices. The code explicitly recognises that creators have established processes for managing the disclosure of their content's origin.

Similarly, text under human editorial control (review + assumption of editorial responsibility) benefits from an adapted regime. Newsrooms using AI as a working tool but where a human validates the final content can rely on their existing procedures.

The timeline: 5 months to prepare

• 5 March 2026 — Publication of the 2nd draft Code of Practice
• 30 March 2026 — Deadline for stakeholder feedback
• June 2026 — Final version of the Code expected
• 2 August 2026 — Transparency obligations come into force (Article 50 AI Act)

The code is voluntary, but it will become the de facto compliance benchmark. Companies that don't align with it risk closer scrutiny from the European AI Office and national authorities.

Does your organisation use generative AI tools to create content? The free AiActo diagnostic identifies in 3 minutes whether you're a provider or deployer under the AI Act — and which transparency obligations apply to you.

Frequently asked questions

Is the AI icon already mandatory?

No. The icon is part of a voluntary Code of Practice, whose final version is expected in June 2026. However, the transparency obligations under Article 50 of the AI Act will be legally binding from 2 August 2026. The code will become the main tool for demonstrating compliance with these obligations.

Is the "fully AI" vs "AI-assisted" distinction definitively abandoned?

In the 2nd draft, yes: the taxonomy has been completely removed. All content under Article 50 will be labelled the same way. Consultation remains open until 30 March 2026, but the trend is clear. The first draft raised too many practical and legal issues to maintain this distinction.

Do I need to label a photo retouched with an AI tool?

If the retouching is limited to standard editing (filter adjustments, brightness, cropping) and does not substantially alter the content, the obligation does not apply — that's the exception provided by Article 50(2). However, if AI was used to generate or substantially manipulate content (element replacement, generation of entire sections), labelling applies.

Will social media platforms have to apply this label?

Platforms that are both deployers (they use generative AI) and online intermediaries (under the Digital Services Act) must combine Article 50 obligations with DSA requirements. They will play a key role in disseminating and preserving labels on AI content shared by their users.

Who decides the final icon design?

A dedicated task force will be formed with the support of Code signatories. It is tasked with developing a harmonised interactive icon at European level. The illustrative examples published in the 2nd draft annex serve as a basis for discussion — the final design has not yet been decided.

The European AI icon is no longer an abstract concept: the first visuals exist, the timeline is set, and the second draft's simplification sends a clear signal. By August 2026, any organisation that produces or publishes AI content in Europe will need to be ready to label it. The question is no longer "if", but "how".