The issue of using copyrighted data for training AI models has been contentious, with several AI companies being sued for copyright infringement. Proposed legislation in Congress could require AI companies to disclose their sources of training data, allowing copyright holders to know if their work has been used without permission. Fairly Trained will not issue certifications to developers that rely on the fair use argument to train their models.
Key takeaways:
- Fairly Trained, founded by former Stability AI vice president Ed Newton-Rex, offers certification to AI companies that prove they have permission to use copyrighted training data.
- The company's first accreditation, the Licensed Model certification, will be given to companies that license protected data to train their models.
- Fairly Trained has already certified nine generative AI companies that work in image, music, and voice generation, including Beatoven.ai, Boomy, BRIA.ai, Endel, LifeScore, Rightsify, SOMMS.AI, Soundful, and Tuney.
- There is proposed legislation currently making its way through Congress that would require AI companies to disclose where they got their training data, allowing copyright holders to know if their work was used without permission.