Fairly Trained offers a certification program — called the Licensed Model (L) Certification — that evaluates generative AI companies to verify their AI models were trained exclusively on licensed or openly licensed data, without relying on unlicensed copyrighted material.

Applicants submit written documentation of their training data sources, including the types of licenses obtained. Fairly Trained reviews the submission, may ask follow-up questions, and grants certification only when confident in a complete picture of data provenance. The certification is re-evaluated annually, and can be revoked if a company's training data practices change. As of mid-2025, the organization has verified roughly 20 small developers spanning image, music, singing voice, and large language model (LLM) categories.