INSCRIPT: EU AI Act

AI is coming to the fore: What has changed since August

AI has long been part of our everyday digital lives – from chatbots on websites to AI-generated images in social media feeds. Under Article 50 of the EU AI Act, which comes into force on 2 August 2026, there will be stricter regulations on when the use of AI must be made transparent to users. This applies, amongst other things, to direct interactions with AI, deepfakes and certain AI-generated or manipulated content.

For businesses, this means the question is no longer just where AI is used, but also when its use must be made visible. This is a good opportunity to take a closer look at websites, content and digital processes – and to establish clear rules on transparency and labelling.

KI muss erkennbar sein

When users communicate with a chatbot or AI assistant, it must be clear from the very first interaction that AI is being used. A reference in the terms and conditions or the privacy policy alone is not sufficient.

Making AI-generated content transparent

If text, images, videos or audio content are created or significantly altered using AI, a disclosure may be required. The disclosure should be clear, visible and placed directly alongside the relevant content.

Clearly label deepfakes

AI-generated content that looks deceptively real requires a particular level of transparency. This includes, for example, manipulated photos, synthetic voices, avatars and AI-generated videos.

Human responsibility remains

AI can support and automate processes. Nevertheless, it should always be clear who is responsible and how users can contact a member of staff – particularly when it comes to support, job applications, quotations or contractual processes.

Taking data protection and documentation into account

The AI Act does not replace the GDPR. Organisations should document which AI systems are in use, what data is processed, and how oversight, audit processes and responsibilities are organised.

Using AI transparently

For businesses, the most important thing right now is to systematically map out their own use of AI. A good place to start is by taking stock: where are AI-generated images, texts or videos already being used? Are there any chatbots or AI assistants? Who checks content before publication, and who is responsible for labelling, approval and accountability? This quickly highlights where specific action is needed.

Building on this, clear processes can be defined: identifying areas of use, establishing labelling for relevant content, integrating notices within chatbots and determining responsibilities for review and approval. It is also important to inform staff, in a concise and understandable manner, about when AI-generated content must be reviewed and labelled. As a digital agency, this is precisely where we can help: with taking stock of existing AI applications, developing a labelling concept, setting up an AI transparency page, providing training for staff, and implementing the technical aspects on websites and within digital processes.

Note: This article is for general information purposes only and does not constitute legal advice.

Artikel 50 legt Transparenzpflichten für bestimmte KI-Anwendungen und KI-generierte Inhalte fest. Nutzer:innen sollen erkennen können, wann sie mit KI interagieren oder wann Inhalte künstlich erzeugt beziehungsweise wesentlich verändert wurden.

Ja. Wenn Nutzer:innen mit einem KI-Assistenten oder Chatbot kommunizieren, muss dies bereits bei der ersten Interaktion klar erkennbar sein. Ein Hinweis nur in der Datenschutzerklärung oder den AGB reicht nicht aus.

Je nach Anwendungsfall kann eine Kennzeichnung erforderlich sein. Besonders relevant ist sie bei realistisch wirkenden oder deutlich durch KI veränderten Bildern, Videos und Audioinhalten.

Täuschend echt wirkende KI-Inhalte wie manipulierte Fotos, synthetische Stimmen, Avatare oder KI-generierte Videos müssen besonders klar gekennzeichnet werden.

Nicht automatisch jeder KI-unterstützte Text. Ob eine Kennzeichnung erforderlich ist, hängt vom konkreten Einsatz und Anwendungsfall ab. Unternehmen sollten deshalb definieren, wie KI-Texte geprüft, freigegeben und veröffentlicht werden.

Eine eigene Transparenzseite ist generell nicht verpflichtend, kann aber sinnvoll sein. Dort lässt sich zentral erklären, welche KI-Systeme eingesetzt werden, wofür sie genutzt werden und wie menschliche Kontrolle sichergestellt wird.