‘It’s high-tech theft’: As AI music booms, Canadian artists demand consent, transparency and pay

Canadian musicians are sounding the alarm over artificial intelligence's unauthorized use of their work to train music-generating algorithms, demanding consent, transparency, and fair compensation. Cadence Weapon, an Edmonton-raised rapper, discovered 129 of his songs in datasets used to train AI music generators, expressing dismay and frustration. He warns that AI inherently relies on scraping existing creative work, creating an existential struggle for artists navigating this new landscape.

The Canadian music industry is pushing for rules that require AI companies to compensate songwriters and musicians whose work powers their systems. Canada's federal government is currently holding a public consultation on AI transparency, aiming to determine what meaningful and practical disclosure should look like as the technology becomes increasingly embedded across the economy. Music organizations like SOCAN and Music Publishers Canada are preparing submissions for the consultation, which runs until September 23.

Artists like Fenix Flexin, Tyga, and Dr. Dre have acknowledged using AI on their music, with the latter claiming it's a tool for creativity. However, independent artists like Pemberton feel squeezed from both sides: their work helps train AI systems, while competing music dilutes payments on streaming platforms. Apple Music and Deezer report that more than a third and 44% of new songs, respectively, are AI-generated, while Spotify claims to have removed 75 million "spammy" AI tracks.

The issue goes beyond whether songs are being used to train models; it's that AI music generators have been unwilling to disclose their training data. Canada's Music Publishers Canada CEO Margaret McGuffin states that the organization frequently hears from songwriters and publishers who discover their music in AI training datasets, believing this to be an infringement of copyright.

Those involved in scraping and training need to pay and obtain consent, she asserts, emphasizing the need for transparency about what has been scraped and used to train models.

Some argue that there are positive uses for AI, but Grammy-winning Canadian roots artist Allison Russell maintains that stealing existing artists' work is not the way to achieve it. She notes that 40 of her songs appear in The Atlantic's database of music identified in AI training datasets. Russell sums up her stance, saying, "They're learning without consent from everything that's out there and then trying to imitate it. We need legal guardrails, frankly. It's high-tech theft."