Lawsuit tracker
Ontario Superior Court of Justice, Toronto · ActiveCanadian News Media v. OpenAI
Canada’s biggest news organizations sued OpenAI together and beat back a jurisdictional challenge — establishing that training a model abroad does not put a company beyond the reach of courts where the content, and the harm, actually sit.
Key facts
- Filed November 29, 2024 in the Ontario Superior Court of Justice by CBC/Radio-Canada, Postmedia, The Globe and Mail, Torstar and The Canadian Press.
- Alleges copyright infringement under Canada’s Copyright Act and breach of website terms of use over scraping Canadian journalism to train ChatGPT.
- Statutory damages under Canadian copyright law can reach C$20,000 per infringed work, on top of disgorgement of profits and an injunction sought.
- OpenAI argued in September 2025 that it has no presence in Ontario and that the case belongs in US courts.
- The court dismissed OpenAI’s jurisdictional challenge on November 27, 2025; the case proceeds in Ontario.
The first Canadian AI-training case
Five of Canada’s largest news organizations — public broadcaster CBC/Radio-Canada, Postmedia, The Globe and Mail, Torstar and the newswire Canadian Press — filed jointly rather than separately, a structure that spreads litigation cost and presents a unified record of alleged scraping across English- and French-language journalism. The claims combine copyright infringement under the Copyright Act with a breach-of-contract theory: that OpenAI violated the terms of use posted on the publishers’ own websites when it scraped their content.
OpenAI’s jurisdictional challenge
Before fighting the merits, OpenAI tried to get the case dismissed on a threshold ground: it argued it has no offices, servers or employees in Ontario, that the alleged scraping and training took place outside Canada, and that any dispute belonged in US courts instead. Canadian courts, like most, require a real and substantial connection between the dispute and the jurisdiction before they will hear a case against a foreign defendant.
Why the court kept the case in Ontario
The Ontario Superior Court of Justice heard the challenge in September 2025 and dismissed it in November 2025, reasoning that the fair and efficient functioning of the Canadian legal system favours letting Canadian creators pursue claims over Canadian-made works, against foreign companies, in Canadian courts. That is a jurisdictional ruling, not a decision on infringement — it does not touch whether OpenAI actually infringed anything. But it removes OpenAI’s easiest exit from the case and confirms the lawsuit proceeds on Canadian soil, under Canadian copyright law.
Fair dealing is narrower than fair use
That distinction matters more here than in similar US suits. Canada’s copyright statute does not have an open-ended fair-use defense; it has fair dealing, a closed list of specific purposes — research, private study, criticism, news reporting and a handful of others — and AI model training does not sit obviously inside any of them. A defendant that would reach for fair use reflexively in a US courtroom has a narrower, more uncertain set of arguments available in Ontario.
Why it matters for training-data licensing
The jurisdictional ruling is itself the headline for data buyers and sellers: a model company with no physical footprint in a country can still be sued there over how it used that country’s content, so long as the harm to local creators is real. Combined with fair dealing’s narrower scope, a merits loss in Canada could land harder on AI companies than an equivalent US ruling would. Any company deploying models trained on scraped content should map its copyright exposure jurisdiction by jurisdiction, not by where its servers happen to sit.
What to watch
- The next procedural steps now that jurisdiction is settled — likely pleadings and early discovery on the merits.
- Whether Canada’s fair-dealing framework gets tested directly against an AI-training defense.
- Whether other Canadian publishers join or file parallel suits following the jurisdictional win.
- Any statutory-damages exposure calculation, given the C$20,000-per-work ceiling across a large journalism catalog.
Sources
See something wrong? Send a correction.
Jaeden Schafer
Jaeden Schafer is the founder of fiund and host of the AI Chat podcast. He covers the training-data market and the lawsuits shaping it.
More case files
Want data that clears this in diligence?
Whether you're building a model or sitting on an archive, the first conversation is short and specific.
Send a brief