EU Turns AI Content Transparency into Operational Requirements

What happened: On July 20, the European Commission published guidelines to help providers and institutional deployers of AI systems apply the AI Act's transparency obligations before they take effect on August 2, 2026. The guidelines explain that providers must notify people when they interact directly with an AI system and add machine-readable markings to generated or manipulated content. Deployer duties also cover disclosure of deepfakes, certain public-interest content published without human review or editorial oversight, and the use of emotion recognition or biometric categorisation systems.

The lens: soft power These details give the European Union the ability to turn its rules into product specifications followed by companies outside its borders when they want access to its market. The power lies not only in owning the model, but in defining the form of the notice and label, and responsibility for publication.

Who is affected: Model and platform providers, compliance and product teams, and media, commercial and government institutions that publish generated content or use conversational interfaces within the European Union.

What it means for the region: Institutions in the Middle East and North Africa that serve European users should not treat transparency as merely a terms and conditions page. In practice, the product lifecycle needs a record of where the notice appears, how content is labelled and who reviews public-interest material before publication.

The practical takeaway: Review three high-volume user journeys in your product this week. For each one, identify when the user is notified, the machine-readable labelling method and the person responsible for human approval, then document the gaps before August 2.

Source: https://digital-strategy.ec.europa.eu/en/news/commission-publishes-guidelines-transparency-obligations-providers-and-deployers-certain-ai-systems