Ed Newton-Rex, founder of Fairly Trained, talks generative AI, copyright and the value of creativity at Music Publishers Canada 2025 AGM

Ed Newton-Rex understands the value of human creativity and wants to make sure AI companies stop scraping the internet for free and fairly compensate creators (and those who invest in them) for using their work in designing and training generative AI models.

Watch the full interview below

It is imperative that AI companies respect Canada’s legally-established copyright protections and enter into licensing agreements that respect the rights of copyright holders.  Canada’s music publishers  also  expect Canada’s AI framework to respect Canadian IP.

During the interview, Newton-Rex told host Lucy van Oldenbarneveld that ultimately, generative AI models need three key resources: people, software engineers to build the models; they need GPUs on which to run the training process, chips, servers; and they need training data, which is people's life work. AI companies spend millions of dollars per software engineer, spend up to a billion dollars per model they create, but at the moment, lots of them in different countries demand access to that third resource — training data, people's life work — for free.

Newton-Rex is a former executive at Stability AI, where he resigned following the company’s admission that they were using copyrighted material without permission. This is a fundamental problem, he said, because not only are creators not being paid for their work but they are often not even asked for permission. On top of this, the generative AI model is openly competing with the human creators that are training them, which could lead to significant job losses.

Some other key takeaways:

Current practices are unethical and illegal: Many AI companies are using copyrighted content without consent, often through large-scale web scraping.

Disclosure and licensing agreements are essential solutions: He advocates for disclosure of what data models are trained on, and licensing of that data from rights holders. Without disclosure, creators and creative industry businesses do not know if their works were used; without licensing, they can’t be compensated.

Generative AI can be a valuable tool for creators and music publishers, but its development must be grounded in permission and fair compensation: As Newton-Rex explains, “None of this would work without the creators themselves,” and AI companies—not individual users—should be held accountable for exploiting creative work. He encourages the use of “fairly trained models” that license all their training data, not just selectively, and wants to see existing laws upheld to ensure creators are paid when their work is used in AI training.