AI Act: transparency
Bricks Technical Documentation
How Bricks meets the transparency obligations of the AI Act for apps built with AI and for AI inside an organisation's processes: content marking, provenance, declared AI assistants, and the limits that remain.
The AI Act's transparency obligations are in its article 50. This page says what Bricks does for them. The other regulations are covered in Compliance evidence for auditors.
Art. 50.2 — Marking content produced by AI
- Every page of a published app says how it was built. Each page carries a machine-readable statement of its construction chain, and its footer shows the EU "AI Modified" icon, linking to a plain-language explanation for the app's users. The organisation cannot remove it.
- Content generated by AI is marked where it appears. A region generated without human review carries the "AI Generated" mark; content written by the organisation or by a visitor is marked as such. Marks use the IPTC vocabulary for the source of digital content.
- Generated images carry their provenance in the file. Images produced by the image model keep their signed content credentials (C2PA) byte for byte, and the stored file is labelled as AI-generated.
- The source of each app carries a provenance manifest: who operates it, how it was built, with which AI assistance, which version and when. It can be read from the source without running the app or contacting Bricks.
Current limits. The provenance manifest is not signed: it discloses how an app was built but does not prove that no one edited it after publication. Text generated by AI is marked only where the app wraps it explicitly.
Art. 50.1 — Telling people they interact with an AI
- Bricks's Studio is an AI studio, and says so throughout. It is named and presented as an AI studio; the person chooses which AI model runs it and where, sets the AI's behaviour, and sees the vitals of each AI turn. Interacting with an AI is evident from that context, which art. 50.1 accepts in place of a separate notice.
- An organisation that puts an AI assistant in front of people, such as a chatbot in its app, must tell them they are talking to an AI. Bricks provides the EU icon for it.
Current limits. An organisation's chatbot shows the notice only where the organisation's app places it.
AI inside an organisation's processes
AI assistants that run in an organisation's workflows are declared with their purpose, the data they receive, the recipient, its jurisdiction, who pays the provider, retention, human review, whether they take automated decisions, and their safeguards. These declarations appear in the organisation's record of processing (GDPR).
Deepfakes (art. 50.4)
Video generation on Bricks is text-to-video only: it accepts no image, clip or voice of a person to modify, and a person, never the AI, starts each render. Bricks therefore does not produce synthetic media of real people.