On September 2, 2026, the Society of Composers, Authors and Music Publishers of Canada, SOCAN , filed a copyright infringement lawsuit against Suno, a US-based AI music generation company, in the Federal Court of Canada. SOCAN is the largest member-owned music rights organization in the country, representing more than 200,000 songwriters, composers, and music publishers. Its lawsuit alleges that Suno built its generative AI models by training on music files freely accessible on the internet, without obtaining permission or paying for a licence, and that the platform has since generated and publicly streamed outputs that closely replicate specific songs already in SOCAN’s repertoire, including Nickelback’s “Photograph” and Avril Lavigne’s “Sk8er Boi.”
The lawsuit itself isn’t unusual on its face; Suno is already fighting several parallel legal battles in the United States, including copyright suits from Universal Music Group, Sony Music Entertainment Group, and Warner Music Group, plus a separate class action filed just a day earlier by singer-songwriter Jason Isbell and other musicians over the alleged unauthorized use of their identities. What makes the Canadian case worth paying attention to is a specific legal difference that could shape how it plays out very differently from its American counterparts.
The defence that works in one country and not the other
In the US lawsuits, Suno has consistently argued that training its models on publicly available material qualifies as fair use under American copyright law, the same broad, flexible, multi-factor defence that Indian courts have also drawn on in adjacent AI training disputes. Fair use in the US is an open-ended balancing test, weighing factors like the purpose of the use, the amount of the original work used, and the effect on the market for the original, which gives a defendant real room to argue that a transformative new use, like training an AI model, shouldn’t count as infringement even without a licence.
Canada does not have fair use. It has a narrower, related doctrine called fair dealing, and the difference between the two is not just terminology, it’s structural. Canadian fair dealing only excuses copying that falls within a closed, specifically enumerated list of purposes set out in the Copyright Act, research, private study, education, parody, satire, criticism, review, and news reporting among them. If a particular use doesn’t fit within one of those named categories, the fair dealing defence generally isn’t available at all, regardless of how transformative or technologically novel the use might otherwise seem. Training a commercial generative AI model isn’t on that list, which means Suno’s core US defence strategy doesn’t automatically translate into a defence under Canadian law the way it might elsewhere.
What SOCAN is actually asking for
SOCAN’s suit centres on Suno’s alleged infringement of performing rights, the right to publicly perform or communicate a musical work, which SOCAN’s members license out and SOCAN itself administers and monitors on their behalf as a collective rights organization. The lawsuit points to a sample of 150 publicly available Suno outputs that SOCAN says closely replicate protected songs in its repertoire, and alongside the court filing, SOCAN has published a website presenting side-by-side comparisons of the original songs and the allegedly infringing AI outputs for public review.
SOCAN is seeking damages, along with an order requiring Suno to stop making the allegedly infringing outputs available and to bring its practices into compliance with Canadian copyright law going forward. SOCAN’s CEO, Jennifer Brown, framed the suit as a matter of principle rather than pure enforcement, stating that “innovation can’t come at the expense of human creativity” and that the organization has “a responsibility to act when the rights of music creators and publishers are put at risk.”
Why this case is worth watching
Generative AI training disputes are being litigated in parallel across multiple countries right now, and they are not all landing in the same place, because they aren’t all being decided under the same copyright framework. A US court weighing Suno’s fair use defence is asking a fundamentally different legal question than a Canadian court will ask under fair dealing, and the answer that satisfies one system’s flexible balancing test won’t necessarily satisfy another system’s closed list of permitted purposes. That’s not a minor technicality; it means the same underlying conduct, training a model on copyrighted material without a licence, can be lawful in one jurisdiction and unlawful in another, decided by nothing more than which country’s copyright statute happens to apply.
For anyone tracking how copyright law is responding to generative AI, this case is a useful reminder that “is AI training fair use” isn’t actually one global question with one eventual global answer. It’s a jurisdiction-by-jurisdiction question, and the specific structure of each country’s exceptions to copyright, open-ended balancing test versus closed enumerated list, will likely end up mattering as much as the underlying facts of what any particular AI company actually did.
Sources consulted for this piece: CBC Music, The Globe and Mail, CP24 News, BNN Bloomberg, the Lethbridge Herald, K-Jewel 99.3 FM, and SOCAN’s own press release announcing the lawsuit. This piece was independently researched and written; no text has been reproduced from any source beyond the short attributed quotes above.
