AI Fair Use Controversy Addressed as Non-profit Certifies Copyright-Free Language Model

Published March 21, 2024

In an industry-shaking move, the new non-profit organization Fairly Trained, founded by ex-Stability AI executive Ed Newton-Rex, has set a precedent by certifying an AI large language model (LLM) as free of copyright infringement. This decision challenges the prevailing belief among AI companies that training such models without infringing on copyrighted materials is infeasible. Ed Newton-Rex, also the founder of the AI music platform Jukedeck, created Fairly Trained to ensure the respect of creators' rights in the AI technology field.

Turning Principles into Practice

Fairly Trained's stance represents a stark contrast to the perspective held by many AI firms, including Stability AI, where Newton-Rex previously worked, and Anthropic, which is currently under legal scrutiny by Universal Music Group for alleged copyright violations. At the core of the disagreement is the principle that when training AI, the effect on the potential market value of copyrighted works should be considered, challenging the 'fair use' argument used by numerous AI companies.

The Path to Ethical AI Development

Just two months after advocating for creators' rights, Fairly Trained has begun its quest by certifying four new AI technologies. The LLM KL3M by 273 Ventures has made headlines as the first of its kind to receive such recognition, indicating that the training of large-scale AI models can abide by ethical standards respecting copyright law. The certification promotes the use of licensed data and acknowledges the importance of human creators in the AI realm.

Expansion and Support for Fair Use AI

The Fairly Trained certification has also been awarded to Voicemod for AI voice and singing models, along with Infinite Album and Lemonaide, two music companies utilizing generative AI, and an AI band named Frostbite Orckings. These certifications mark a significant step forward, demonstrating that fair training practices are achievable across various creative fields.

In support of these movements, influential organizations like The Authors Guild, SAG-AFTRA, Music Managers Forum–US, and the American Society for Collective Rights Licensing (ASCRL) have thrown their weight behind Fairly Trained's initiative. These groups criticize AI companies for exploiting copyrighted work without due consent and underscore the value of the certification in promoting ethical AI development.

ethical, AI, copyright