Fairly Trained

Fairly Trained is a certification program for models and model builders that claim to use properly licensed or consented training data. The program turns sourcing norms into a concrete third-party signal that partners, customers, and policymakers can evaluate, even if announcement-based certification counts still understate the full set of participating models.

Fairly Trained evaluates data sourcing or licensing practices through certification for multimodal material at the training stage. Certification communicates the result of a review or attestation; it does not itself authorize, prevent, or enforce a use. Public materials describe a currently available initiative; the newest dated source in this profile is “Certification program continues to accept applications” (July 16, 2026). These details describe the published mechanism and evidence, not a finding about legal validity, adoption, or effectiveness.