New nonprofit to certify AI models that only use copyrighted training data with permission

A new nonprofit organization, Fairly Trained, has been launched to certify AI models that only use copyrighted training data with permission. The organization, founded by Ed Newton-Rex, former VP of audio at Stability AI Ltd, will issue certifications to AI companies that do not include copyrighted materials in their training datasets unless they have obtained permission. The certification process requires AI developers to implement a due diligence workflow for ensuring the training data they collect is not subject to copyright restrictions and to create a database that logs the records used in each AI training project.

Fairly Trained's certification costs $500 for companies with over $10 million in annual revenue, with a $6,000 yearly fee if the application is accepted. AI developers with annual revenues of less than $10 million can obtain a certification for as little as $500 per year. The nonprofit has already issued certifications to nine generative AI startups and plans to roll out additional certifications in the future, covering more aspects of AI developers’ training data sourcing practices.

Key takeaways:

A new nonprofit, Fairly Trained, will certify AI models that only use copyrighted training data with permission.
The organization was founded by Ed Newton-Rex, former vice president of audio at Stability AI Ltd, in response to the use of copyrighted materials in AI training datasets.
AI developers can obtain a certification from Fairly Trained by implementing a due diligence workflow and creating a database that logs the records used in each AI training project.
Fairly Trained has already issued certifications to nine generative AI startups and plans to roll out additional certifications in the future.

New nonprofit to certify AI models that only use copyrighted training data with permission - SiliconANGLE

Key takeaways:

Comments (0)

Newsletter