Transparency is the obligation most businesses will meet first under the EU AI Act. It is not about proving your AI is safe — it is about being honest that AI is involved. This guide covers what you must disclose and how.
Disclosing AI interaction
When a person interacts with an AI system — a chatbot, a voice assistant, an AI agent — they must be told, clearly and at first contact, unless it is already obvious. A simple, accessible line such as “You’re chatting with an AI assistant” satisfies this for most websites. The disclosure should be perceivable to everyone, which is where accessibility (WCAG) and the AI Act overlap.
Labelling AI-generated content
Synthetic media — AI-generated images, audio, video and, in defined cases, text — must be marked as artificially generated. Providers of generative systems carry the machine-readable marking duty; deployers who publish deep fakes or AI-generated public-interest text carry a visible disclosure duty. In practice that means a label users can see and, increasingly, metadata machines can read.
Keeping proof
Transparency is only worth anything if you can show it. Capture the disclosure each system presents, record where it is used, and date the record. When a customer’s procurement team asks you to “prove you comply with the AI Act”, that dated inventory is the answer — and it is far easier to maintain continuously than to reconstruct under pressure.