Mimikama
Examples EU icons DEEN to the check →
AI Act check

Model statement for your website

Please read first

A general statement about the use of AI on a website is not required under Article 50 of the EU AI Act. This model is a voluntary transparency measure. It does not replace a disclosure that is required directly at a particular AI system or piece of content. A deepfake subject to disclosure must be disclosed at the image or video – a note in the imprint is not sufficient for that.

Depending on your sector and the processing involved, data protection law, copyright, media law, competition law or specific statutory information duties may apply in addition. This model is editorial guidance, not legal advice.

Two places need filling in

The model contains two placeholders in square brackets: the person or role holding editorial responsibility, and the date. Both are highlighted below.

Copy from here

How we use artificial intelligence

We use tools based on artificial intelligence in selected areas of our work. Depending on the task, these systems may assist with structuring information, drafting text, translation, transcription, image editing or the creation of illustrations.

Responsibility

AI systems support our work. Decisions and publications remain the responsibility of people. Editorial responsibility for published content is held by [insert name or role].

Content is reviewed before publication in accordance with its purpose. Checking the accuracy of the content is the minimum requirement. Beyond that, we verify sources, figures, names, context and image content. Reviewing spelling or formatting alone is not something we regard as substantive review.

We do not publish texts on matters of public interest fully automatically and without appropriate human or editorial control.

After editorial sign-off, no AI system intervenes substantively in a text – not for automatically generated headlines, standfirsts or summaries either. Should a substantive AI intervention nonetheless occur after sign-off, the amended final version is reviewed and approved again. Where that review does not take place, we disclose the text accordingly.

Disclosure of AI content

Realistic image, audio or video content that has been generated or substantially manipulated with AI, and that could be mistaken for genuine recordings, is disclosed by us directly at the content concerned. Depending on the depiction we use notes such as “AI-generated depiction”, “image created with AI”, “image altered using AI”, “video contains AI-generated scenes” or “voice generated with AI”.

Ordinary technical editing that does not substantially change the content or meaning of an image, video or text is not disclosed separately. This includes adjustments to exposure and colour, noise reduction, cropping or measures to improve accessibility. We do not remove existing disclosures or machine-readable markings in third-party content.

Communication with AI systems

Where we use systems that communicate directly with people, we indicate at the point of interaction that an AI system is involved, unless this is already unmistakably clear from the context.

Data protection

When using AI systems we take account of the protection of personal and confidential information. Sensitive data is processed only where an appropriate legal and technical basis exists.

Updates

We review our approach to artificial intelligence regularly and adapt this statement to legal, technical and organisational developments.

Last updated: [insert date]

End of the model statement

Where to put it

The statement belongs on a page of its own, for instance under the heading “How we use artificial intelligence”. Link to it where people would expect to find it: in the footer, in the imprint or in your privacy notice.

It is not sufficient where a specific obligation applies. A deepfake needs the disclosure at the image itself; an unreviewed AI text on a matter of public interest needs the notice at the text. Whether that applies to you is answered by the check in a few questions.

This model may be used, adapted and published freely. Attribution is not required, though it is always welcome.

This model statement offers journalistic orientation on the EU AI Act and does not replace individual legal advice. What governs is the specific case, the regulation as it applies and its interpretation.