EU AI Act

Regulation (EU) 2024/1689 follows a risk-based approach: the higher the risk of an AI system, the stricter the requirements — from prohibited practices and high-risk systems down to transparency obligations for everyday applications. Since 2 August 2026, the transparency obligations of Article 50 apply. They cover in particular deceptively realistic deepfakes and AI-generated texts that inform the public about matters of public interest. The European Commission provides three uniform icons for this purpose — AI, AI GENERATED, and AI MODIFIED. Using these icons is voluntary; the legally required disclosure itself, in the cases it covers, is not.

In day-to-day business, this means not every use of AI triggers a labelling obligation. For texts on matters of public interest, the disclosure obligation does not apply where a human reviews the content and a natural or legal person takes on editorial responsibility. For small and medium-sized businesses, a clear, documented process is therefore the most sensible first step: who reviews AI-generated content before publication, who decides on a disclosure, and who holds editorial responsibility? Whether a specific case must be labelled depends on the concrete use case and is no substitute for legal review.

Chifty is built around workflows in which people stay in control: Chifty Office prepares email replies as drafts but never sends automatically — every draft is reviewed and sent by a human. Answers from the knowledge base and in chat show their sources, so it stays traceable what a statement is based on. Throughout, one principle applies: the AI must not bypass permissions — it only draws on sources the asking person is allowed to see. This human review step and the visible sources support exactly the kind of editorial control the EU AI Act's transparency obligations are concerned with.

Transparency at Chifty also means that the chat shows, for every answer, which AI model was selected for it — and that AI processing takes place in the EU by default.