AI Act disclosure for a synthetic construction photo shown to a client
Label synthetic construction imagery clearly, preserve its purpose and human approval, and avoid presenting a concept image as site evidence.
A synthetic renovation image can help a client discuss options, but it can also be mistaken for a real site photo, completed detail or measured design. The risk is not solved by adding “AI” somewhere in a footer. The disclosure must reach the viewer when the image is first presented and remain attached when it is reused.
Start with the current official rule
The European Commission's Article 50 transparency guidelines explain obligations applying from 2 August 2026 for certain AI-generated or manipulated content. Scope depends on the system, content and use. Use the current guideline and legal review for the exact case; do not label every edited construction image a deepfake or assume that a generic disclaimer covers all channels.
Turn the rule into a bounded decision
Classify the asset as site evidence, design visualisation, marketing content or internal draft. If it is synthetic or materially manipulated, state that clearly next to the image and describe what is illustrative. Keep dimensions, specifications and contractual promises in verified documents. A client approval of a concept does not turn generated pixels into an as-built record.
Build an auditable workflow
Store the source files, generation or edit date, purpose, reviewer and approved caption. Require human review before email, proposal or project-message use. When the image moves to another document, carry the disclosure with it. Remove obsolete drafts so a later project team cannot mistake them for photographic evidence.
Handle exceptions explicitly
Test the workflow with a realistic exception: a missing company identifier, duplicate record, unclear role or temporarily unavailable source. The process should stop with a focused question. Keep the source, review date, owner and decision without copying more personal data than needed. Escalate legal, tax, labour or safety uncertainty to the competent person.
Repeat the check when a payment, party, invoice action, site arrangement or official instruction changes. An earlier result is context, not automatic permission for the next action. Define who can reverse an error, notify affected people and decide that the exception is genuinely closed.
Make the status readable to someone who did not prepare the record. That person should find the checked input, consulted source, remaining uncertainty and next permitted action. Avoid relying on a free-text note as the only evidence. Use stable fields for identity, date, outcome and owner, then add only the explanation needed for the exception.
Connect context to controlled execution
Enfin can connect the customer, supplier, project, document, task and invoice in one record. Alfie can prepare information and propose a governed action within available permissions; the official service and competent owner remain responsible for the decision. Explore Enfin, review pricing, or discuss your workflow.
AI Act disclosure for a construction chatbot: identify the system and preserve human handoff
Design a construction chatbot disclosure that identifies AI use, preserves human handoff and keeps project decisions under accountable review.
Add a customer note from a WhatsApp photo with Alfie and preserve its source
Attach only the relevant photo context to the correct customer and project while keeping identity, permissions and follow-up explicit.
Construction client portals: fewer status calls, more trust
Construction client portals: fewer status calls, more trust. Compare workflows, data and implementation across Belgian construction.