page / MENA · AFRICA

Human data should arrive with a clear history.

The origin of data, the permission to use it and the conditions under which people contributed are part of dataset quality.

01Consent

Explain the work and the intended use.

Participants should receive clear information about the task, capture method, data categories, intended AI use, retention and available withdrawal or escalation routes where applicable.

02Contributor standards

Treat human judgment as skilled work.

Instructions, compensation, feedback routes and acceptance decisions should reflect task difficulty and local conditions. Completed work should not be rejected without a review path.

03Provenance

Keep origin and permission attached.

Delivery evidence should connect each batch to its protocol, consent version, collection context, permitted use, transformations and review history.

Tell us what your model needs to understand.

Discuss a Data Program