- Identify each system’s role and audience.
- Provide information when it actually helps the user.
- Retain classification decisions and acceptance evidence.
The 2026 context
On 20 July 2026, the European Commission published guidelines on transparency obligations under Article 50 of the AI Act. The page, updated on 31 July, states that these obligations apply from 2 August 2026. Scope depends on the system, its use and the organisation’s role.
For a Moroccan business working with European clients, the first decision is to have relevant activities assessed. This framework prepares a project; it is not legal advice. It helps gather facts for that assessment and turn the resulting decisions into product behaviour.
Map where AI intervenes
Walk through the service as a customer: opening a chatbot, reading a reply, downloading a document, viewing an image or transferring to an adviser. At each step, record what AI produces and what the user might reasonably understand. A single product can contain several different situations.
Include content prepared internally and then distributed elsewhere. Assign an inventory owner and a review frequency. A new feature or use should trigger reassessment of the relevant entry. This prevents transparency information from remaining static while the service changes.
Write useful, timely information
For a demonstration assistant, wording such as “AI assistant — responses may need checking” gives immediate context. Then explain available sources, scope limits and how to reach a person. Exact wording must reflect the assessment and actual service behaviour.
Place information in the main journey, with readable contrast and keyboard and screen-reader access. Check exports and translations too. Information that is clear on desktop can disappear in a mobile view or shared content.
Make verification observable
Prepare acceptance checks with screenshots, versions and owners: first contact, resumed conversation, system error and human handover. Test missing-source and unable-to-answer cases. Users should know how to continue towards their goal.
Retain the rationale for chosen notices and the date reference documents were checked. At the next change, the team can compare expected behaviour with existing evidence. This record supports product maintenance and collaboration between business, design, engineering and legal advisers.
| Situation | Question to document | Acceptance evidence |
|---|---|---|
| Interaction | Does the user understand what they are interacting with? | First screen, keyboard, mobile |
| Distributed content | What information should accompany this content? | Published and exported versions |
| Limit or error | How can the user get human help? | Tested handover journey |
Sources & methodology
This insight combines cited publications with editorial analysis. Illustrative examples are not client results. Vendor features and terms may change.