New EU Guidance on AI Transparency: What Should Companies Be Doing From 2 August 2026

This blog from authors Elisabet Cots, Silvia Sáenz de Ormijana & Rut Alemany (Baker McKenzie) offers an overview of the European Commission’s draft Guidelines on the AI Act’s transparency obligations under Article 50, alongside a new Code of Practice for marking and labelling AI-generated content. Together, these documents provide the clearest indication yet of how regulators are expected to interpret and enforce AI transparency requirements before they take effect on 2 August 2026.

The authors explain that the draft Guidelines introduce a practical four-part test for determining whether AI-generated or manipulated content qualifies as a “deepfake.” The assessment focuses on realism, plausibility, references to real or fictional subjects, and whether audiences could perceive the content as authentic, regardless of any intent to deceive. The guidance also clarifies that while minor AI edits are generally excluded, more substantial alterations are likely to trigger disclosure obligations.

The post also examines the accompanying Code of Practice, which provides practical recommendations for providers and deployers of AI systems. It covers the labelling of AI-generated and manipulated content, user notifications when interacting with AI systems such as chatbots, and machine-readable markings that enable AI-generated content to be identified. Companies that adopt the Code, once formally endorsed, are expected to benefit from greater legal certainty in demonstrating compliance.

The authors conclude that businesses should begin preparing now by reviewing how AI is used across their content workflows, particularly in marketing and customer-facing communications. Beyond meeting legal requirements, they argue that transparent AI practices will become increasingly important for maintaining consumer trust, protecting brand reputation, and reducing future regulatory and enforcement risks.