Practical answers to AI-content transparency questions.
Source-based explanations of Article 50, the voluntary Transparency Code of Practice and the publishing decisions organisations need to make before 2 August 2026.
What is Article 50 of the EU AI Act?
A structured explanation of who must comply, which systems and content are covered, the difference between provider and deployer duties, the human-review exception and the 2026 application dates.
Read the full guideStart with the question your organisation is facing.
Each page gives a direct answer, examples, boundaries and primary EU sources.
What is Article 50?
The full map of interaction notices, machine-readable marking, deepfake disclosure and public-interest text.
Read the guide → CodeIs the Code of Practice mandatory?
Why the Code is voluntary while applicable Article 50 obligations are legally binding.
Read the guide → WorkflowWhat does human review mean?
How substantive review, authority to change content and editorial responsibility fit together.
Read the guide → Risk & evidenceWhy do transparency records matter?
The legal penalty framework and the practical costs of weak documentation.
Read the guide →Next questions in the Article 50 series.
The knowledge base will expand around specific search questions rather than general AI commentary.
- 02Does all AI-generated content need to be labelled?Scope, content type and context.
- 03What is the difference between a provider and a deployer?Roles, examples and mixed situations.
- 04What counts as AI-generated versus AI-assisted?Standard editing, substantial alteration and practical classification.
- 05What counts as a deepfake under the AI Act?The legal definition and disclosure boundary.
Trust starts with showing the source and the limit.
Primary sources first. The regulation, Commission material and the AI Act Service Desk take priority.
Status is explicit. Binding law, voluntary codes, draft guidance and ShowAI recommendations are kept separate.
Updates are dated. Every legal guide shows when it was last substantively reviewed.
No false certainty. Unsettled interpretation is identified rather than hidden.