On 31 July 2026 the European Commission announced that, from 2 August 2026, the Commission’s AI Office and national authorities begin enforcing the Artificial Intelligence Act, and that a new set of transparency requirements starts to apply on the same date. This is the point at which the AI Act moves from a published rulebook with staggered deadlines into an actively supervised regime, with an identified enforcement apparatus behind it.
The transparency obligations taking effect are aimed at disclosure rather than product design. Chatbots and other interactive AI systems have to tell users they are dealing with AI and not a human. Deepfakes - manipulated or synthetically created media - must be labelled as such. AI-generated or AI-altered content must carry machine-readable marks so that it can be detected automatically downstream. The Commission frames the package as reducing deception and manipulation while giving businesses clearer obligations and a practical route to demonstrate compliance. It also noted that more than 180 organisations have already signed the Code of Practice on transparency of AI-generated content, the voluntary instrument that maps these duties into concrete practice.
Alongside enforcement, the Commission pointed to the channels through which problems will reach it: an AI Act complaints tool, an AI Act Whistleblower Tool, and a dedicated complaints channel for downstream providers that build on general-purpose AI models. That last channel matters commercially, because it gives companies that deploy someone else’s foundation model a formal route to raise issues with the model provider’s conduct.
For a business leader deploying AI in the EU, the practical read is that user-facing disclosure is now a live compliance surface, not a future one. Any customer-facing chatbot, any synthetic image or video used in marketing, and any AI-assisted content pipeline needs to be checked for whether the AI involvement is disclosed and whether generated output carries machine-readable provenance marking. Note that this date is separate from the high-risk system obligations, whose timing was moved by the Digital Omnibus process; the transparency and general-purpose AI supervision elements are what land on 2 August 2026.