Our first certification is the Licensed Model (L) Certification for AI providers. It can be awarded to any company, organisation, or product that makes generative AI models or services available, and that meets the requirements below.

L Certification requirements

1. What can be certified?

You can apply for certification for an entire company, a single product or service, or an individual model that makes up part of your product or AI architecture.

2. Data sources

All of the training data used for the model(s) being certified must fall into one of the following categories:

• Be explicitly provided to the model developer for the purposes of being used as training data, according to a contractual agreement with a party that has the rights required to enter such an agreement

• Be available under an open license appropriate to the use-case

• Be in the public domain globally

• Be fully owned by the model developer

Obtaining a license from an organisation that itself licenses from creators (e.g. a record label or a stock image library) is considered consent for these purposes. (For more info on this, see our FAQs.)

Any third-party models, open models or synthetic data you use in the product / service / models being certified must adhere to the same requirements.

3. Data due diligence

You must have a robust process for conducting due diligence into the training data you consider using, including thorough checks into the rights position of the training data provider.

4. Record keeping

You must have a robust process for keeping records of the training data that was used for each model training.

5. Further requirements for applicants for individual model certification

Applicants for individual model certification are required to disclose publicly which of the models they use are and are not certified. This is done using consumer-friendly language in your website FAQ, agreed upon with Fairly Trained in advance of certification completion.