A registration nobody scrutinized, five years later
In December 2021, the Canadian Intellectual Property Office (CIPO) registered copyright in Suryast, a sunset photograph run through an AI style-transfer tool built by its own creator, Ankit Sahni, to mimic Vincent van Gogh's The Starry Night. The registration listed Sahni and the AI tool, RAGHAV, as co-authors — apparently the first time a Canadian copyright record has named a machine as an author. That was possible because CIPO does not examine copyright claims on the merits; an applicant fills out a form and pays a modest fee, and the registration issues on its face (The Globe and Mail).
The registration sat unchallenged until the Samuelson-Glushko Canadian Internet Policy and Public Interest Clinic (CIPPIC) filed a Federal Court application in 2024, court file T-1717-24, to expunge or amend it (CIPPIC v Sahni). CIPPIC's case rests on two claims: that Suryast lacks the "skill and judgment" the Supreme Court has said originality requires, since an AI system cannot exercise human faculties, and that Sahni's role — selecting a photo and a style reference — was "trivial and mechanical," not authorship. Sahni disagrees, arguing the Copyright Act nowhere bars AI involvement and that his creative choices, plus RAGHAV's transformation, were both "essential and unique" contributions.
That two-sided fight just gained a third voice. Music Publishers Canada was granted leave to intervene in June 2026 and filed its submissions on July 15, 2026 — the news hook for this piece (CelebrityAccess).
What the music publishers actually argued
Music Publishers Canada's position is narrower than a flat anti-AI stance. It told the court that "only a human can be an author for copyright purposes," but paired that with an explicit case-by-case standard: courts should assess "the extent of the human author's contributions of skill and judgment" in each work, rather than banning AI-assisted registration outright. The publishers also asked the Federal Court to keep Canadian doctrine aligned with the direction other jurisdictions are taking on AI-assisted authorship (CelebrityAccess; Billboard Canada).
That framing deserves to be taken on its own terms before it's argued against. Music publishers have concrete reasons to fear a permissive rule: if a co-authorship claim can attach to whatever an AI tool outputs from a human prompt, the practical effect is to dilute the value of registrations that actually reflect a composer's or lyricist's judgment, and to invite a wave of thin, AI-assisted claims crowding a system that was never built to examine them. Canada's own consultation record backs the sympathy for that worry — in its 2025 "What We Heard" report on copyright and generative AI, the federal government found that consultation participants broadly wanted human authorship to "remain central to copyright protection," with support pointed specifically at requiring "sufficient human contributions" before AI-assisted work qualifies (ISED). That is not a fringe position; it is where the country's own stakeholders, including creative industries, already sit.
Where the case-by-case standard undersells the problem
But Music Publishers Canada's own preferred remedy — a contextual, skill-and-judgment inquiry applied case by case — is close to what the law already does for human creators, and that's the point CIPPIC's core argument gets right that a purely contextual test risks obscuring: the question isn't how much judgment went into a work, it's whether a non-human entity can hold an authorship interest at all. CIPO's practice of registering AI as a co-author, rather than treating the human operator as the sole author of an AI-assisted work, is the actual defect worth fixing, and it's a narrower, more administrable fix than either side's full position. The U.S. Copyright Office reached exactly that outcome on this same image, rejecting Sahni's American filing three separate times on the grounds that human and machine contributions to Suryast "could not be untangled" and that the registrable elements originated from the machine, not the applicant (The Globe and Mail). Canada doesn't need to import that reasoning wholesale, but the discrepancy between an unexamined $63 Canadian filing and three considered U.S. rejections of the identical work is itself evidence that CIPO's intake process, not just Federal Court doctrine, needs tightening.
Why the outcome matters beyond one sunset photo
A Federal Court ruling that AI tools cannot be named co-authors would not, by itself, chill Canadian AI investment — it would simply confirm what CIPO's own registration practice should already reflect and what most consultation respondents told the government they wanted. The risk to innovation runs the other way: if courts instead entrench an ambiguous co-authorship category, every Canadian startup building on generative tools inherits legal uncertainty about who owns the output, discouraging exactly the commercialization the sector needs. A clean human-authorship rule, applied with Music Publishers Canada's sensible reminder that courts should still look at the extent of genuine human creative input, gives AI-assisted creators a predictable path to protection without pretending a model is a legal person. The Federal Court has not yet set a hearing date, but with three parties and a formal government consultation record now pointing the same direction, Canada is closer than most jurisdictions to resolving the question cleanly.