By Tom Janssens · Founder, Easy Cyber Protection · Last reviewed 12 June 2026 · How we write this

AI-generated content: the EU labelling rules explained

If your business uses artificial intelligence (AI) to help write blog posts, social media updates, product images, or video, new European Union (EU) transparency rules concern you. From 2 August 2026, Article 50 of the EU Artificial Intelligence Act (AI Act) requires that certain AI-generated content is clearly identifiable. This guide explains what that means for a small and medium-sized enterprise (SME), in plain language.

Magnifying glass over a published web article revealing a clearly visible AI-generated content label — From 2 August 2026, some AI-generated content must carry a visible label

What Article 50 requires, from 2 August 2026

Article 50 of the EU AI Act contains the transparency obligations. They apply from 2 August 2026. In plain language, three duties matter for most businesses:

→ Chatbots must be recognisable. When people interact with an AI system, such as a chatbot on your website, it must be clear they are talking to a machine, unless that is already obvious from the context.
→ AI-generated content must be machine-readably marked. Providers of AI tools must mark AI-generated or AI-manipulated content (text, images, audio, video) in a format that software can detect. This duty sits with the tool makers, not with you as a user.
→ Deepfakes and public-interest text must be visibly labelled. Businesses that publish deepfakes, or AI-generated text published to inform the public on matters of public interest, must disclose clearly that the content was artificially generated or manipulated. This duty sits with the publisher.

Provider or deployer: which one are you?

The AI Act splits the obligations between two roles. A provider builds or supplies the AI system: the companies behind the chatbots and image generators. A deployer uses an AI system under its own authority: that is you, the moment your business uses AI tools in its work. An SME publishing AI-assisted marketing is a deployer. The practical consequence: the machine-readable marking is your tool vendor's job; the visible labelling of deepfakes and public-interest text, and being transparent about chatbots on your own website, is yours.

The Code of Practice of 10 June 2026: the Commission's own how-to

On 10 June 2026, the European Commission published the final Code of Practice on marking and labelling AI-generated content (source: digital-strategy.ec.europa.eu). It is voluntary: you are not legally required to follow it. But it is the Commission's own description of how to mark and label correctly, which makes it the low-risk path. If you follow the Code, you have a strong answer to the question "did you comply with Article 50?". For an SME the takeaway is simple: when you need to label something, do it the way the Code describes, visibly and unambiguously, rather than inventing your own format.

The liability backdrop: AI output counts as your own words

Labelling is only half the story; responsibility is the other half. On 28 May 2026, the Munich Regional Court decided in case 26 O 869/26 (reported between 9 and 11 June 2026 by the-decoder.com) that Google's AI Overviews are Google's own content. The court issued a preliminary injunction after the AI summaries spread false scam claims about two Munich publishers. The principle the ruling establishes: AI output is the publisher's responsibility, not a neutral conduit. It is a German preliminary decision, not Belgian case law, but the direction is clear. If your business publishes an AI-drafted text containing a false claim, that claim is yours. A label does not remove that responsibility; only human review does.

Practical steps for an SME

You do not need a legal department for this. Three habits cover most of the risk:

1. Know where AI writes for you

List the places where AI-generated or AI-assisted content reaches the outside world: website texts, blog posts, social media, newsletters, product images, video, and any chatbot on your site. You cannot label what you have not mapped.

2. Label where required

Deepfakes and AI-generated text published on matters of public interest must carry a clear label from 2 August 2026. A chatbot on your website must be recognisable as a chatbot. For other AI-assisted content a label is not mandatory, but a short disclosure builds trust.

3. Keep human review on everything you publish

A person reads, corrects, and approves before publication. This is your defence against the liability risk from the Munich ruling, and it catches the factual errors AI tools still make.

Put these three habits in writing and you are most of the way to an acceptable-AI-use policy. Our guide Acceptable AI use at work turns them into a full one-page policy.

The MSP angle: have a standard answer ready

If you are a managed service provider (MSP), your clients will ask "do we need to label this?" the moment the rules start applying on 2 August 2026. Prepare a standard answer instead of improvising per client:

→ A one-paragraph explanation of who is a provider and who is a deployer, so the client understands which duties are theirs.
→ A template disclosure line clients can reuse, for example: "This content was created with AI assistance and reviewed by our team."
→ A short checklist per client: where does AI publish for them, is there a chatbot, who does the human review?
→ A pointer to the Commission's Code of Practice of 10 June 2026 as the reference for how to mark and label.

FAQ

Does every AI-assisted text need a label?

No. The visible labelling duty for businesses targets deepfakes and AI-generated text published to inform the public on matters of public interest. Ordinary product pages or marketing copy fall outside that strict duty. A voluntary disclosure is still a good idea: it costs one sentence and builds trust.

We only use AI for internal documents. Do the rules apply to us?

The labelling duties are about content you make public and about AI systems people interact with, such as chatbots. Internal drafts and analyses are not the target. The moment AI-assisted content leaves the company, the question becomes relevant.

Is the Code of Practice of 10 June 2026 mandatory?

No. It is voluntary. But it is the European Commission's own description of how to mark and label AI-generated content correctly, so following it is the lowest-risk way to show you took Article 50 seriously.

What does the Munich ruling mean for my business?

The Munich Regional Court treated AI-generated output as the publisher's own content in a preliminary injunction of 28 May 2026 (case 26 O 869/26). It is a German preliminary decision, not Belgian precedent, but the message travels: if you publish AI output, you answer for it as if you wrote it yourself. Human review before publication is your protection.