Ethical AI Certification Program Emerges to Tackle Copyright Concerns

In a significant development within the rapidly evolving landscape of artificial intelligence (AI), a new certification program, Fairly Trained, has emerged, addressing the contentious issue of copyright violations in AI models. The initiative comes in the wake of legal battles and growing concerns surrounding the use of copyrighted data in training generative AI models.

Fairly Trained: A Response to Exploitation Concerns

Fairly Trained, founded by former Stability AI vice president for audio, Ed Newton-Rex, has introduced a certification program to recognize and endorse AI companies that proactively seek permission to use copyrighted training data. Newton-Rex, who left Stability AI in November, citing concerns about generative AI exploiting creators, aims to set ethical standards within the industry through Fairly Trained.

Licensed Model Certification: A Benchmark for Ethical AI

The flagship accreditation, dubbed the Licensed Model certification, is awarded to companies that can demonstrate their commitment to ethical AI practices by licensing protected data for training their models. Fairly Trained explicitly states that it will not issue certification to developers relying on the fair use argument, emphasizing the importance of explicit permission when using copyrighted material.

To date, Fairly Trained has certified nine generative AI companies operating in various domains, including image, music, and voice generation. The certified companies include Beatoven.ai, Boomy, BRIA.ai, Endel, LifeScore, Rightsify, SOMMS.AI, Soundful, and Tuney.

Addressing Legal Challenges in Generative AI

The certification program comes amid a surge in legal challenges related to copyright infringement in generative AI. Artists and authors have filed lawsuits against several AI companies, alleging unauthorized use of their copyrighted works. Notably, The New York Times has taken legal action against OpenAI and Microsoft, accusing them of violating its copyright during the training of GPT models.

Proposed Legislation: A Call for Transparency

In response to the escalating concerns, Congress is considering legislation that would mandate AI companies to disclose the sources of their training data. This proposed regulation aims to provide copyright holders with information about the use of their work, ensuring transparency and accountability within the AI industry.

Here is a summary table of the certified generative AI companies:

As AI technology continues to advance, initiatives like Fairly Trained and legislative efforts are crucial in establishing a responsible and ethical framework for AI development. These endeavors seek to strike a balance between fostering innovation and respecting intellectual property rights, ensuring a sustainable and ethically sound future for AI in South Africa and beyond.