AI labelling becomes mandatory in August: what the new EU code means for your business

From 2 August 2026, the transparency obligations of the EU AI Act (Article 50) become binding. Anyone operating a chatbot or producing content with generative AI — which today includes a large share of mid-sized companies — will have to label it. Since this week there is a practical template: the European Commission has published the final code of practice on marking AI-generated content.

What applies from 2 August

Article 50 is not aimed at tech corporations alone — it applies to companies that use AI in everyday customer-facing work. The core obligations:

• Chatbots must make clear that users are talking to an AI
• AI-generated content (text, image, audio, video) must carry a machine-readable marking
• Deepfakes must be visibly labelled
• AI-written texts on matters of public interest require a disclosure notice

The code of practice: voluntary, but useful

The code itself is voluntary — the obligations behind it are not. Its value lies in the detail: it translates the legal text into concrete practice, including a common EU icon set for labelling. Companies that follow it gain planning certainty and a simple way to demonstrate compliance.

Labelling is not a formality — it is trust work. Customers reward transparency, and regulators expect it anyway.

What to do now

• Take stock: where does AI-generated content appear in your company, where do chatbots run?
• Check whether your tools support machine-readable marking (e.g. watermarking, content credentials)
• Add clear labelling to your website, chatbot and content workflows
• Assign responsibility and use the code of practice as your checklist

Started early, the effort is manageable — most of it is a one-time review of tools and processes. If you would like support assessing your AI setup, we are happy to help.

Explore our AI solutions