Transparency that survives the pipeline
Article 50 of the EU AI Act sets transparency expectations for generative AI. Meeting them is not a policy question, it is a workflow question: whether the provenance signals you intend actually reach the people who see the content. The Transparency Workflow Review takes the generative workflows you run and the artefacts you publish, and tests them end to end.
What we check
Provenance metadata, watermarks and labels on the content you generate.
Whether those signals survive the publishing pipeline, from generation to what the audience finally receives.
How synthetic imagery is produced, labelled and released. Synthetic imagery is covered here, not as a separate service.
Where the workflow drops a signal, and what that leaves exposed.
Scope and timing
Two weeks. It can be commissioned on its own, or added to a larger engagement. It is an advisory review and guidance, reported in the language of alignment risk, potential exposure and evidence gaps.
Where it sits
This is a supporting line, not the full picture. It is a common way for smaller organisations to start, and it can precede the full Assessment or follow it as focused work on a specific exposure. The Assessment remains the entity-level review of how the company governs its AI as a whole.