Ontario’s Court of Appeal ruled Friday, November 13, that OpenAI must defend a copyright lawsuit from five Canadian news organizations in an Ontario courtroom, where the U.S. fair use doctrine does not apply and the Copyright Act contains no text-and-data-mining exception for AI training. The Court of Appeal dismissed OpenAI’s appeal, according to The Canadian Press.
The ruling upholds a November 2025 jurisdiction decision by Justice Kimmel of the Superior Court, which found that Ontario courts can hear claims that OpenAI scraped and copied content from Torstar, Postmedia, The Globe and Mail, The Canadian Press and CBC/Radio-Canada to train the models behind ChatGPT. OpenAI had argued the dispute belongs in a U.S. court.
Justice B.W. Miller wrote the unanimous decision rejecting OpenAI’s appeal, and Broadcast Dialogue reported the panel’s finding that “Ontario has a real and substantial connection to the alleged breaches.” The same outlet reported that OpenAI must pay the publishers $100,000 in costs; it did not name a currency, though Ontario courts typically award costs in Canadian dollars.
The appeal court also noted that OpenAI filed no evidence showing a Canadian judgment would conflict with U.S. law or complicate enforcement, and it relied on the efficient functioning of the Canadian legal system, which favors letting Canadian claimants sue foreign companies for breaches of Canadian law, The Canadian Press reported. The Canadian Press is a plaintiff in the case.
Appeal Court Left Open Whether OpenAI’s Training Happened in Canada
The appeal decided where the case will be heard, not whether OpenAI infringed anyone’s copyright. The most consequential question still unanswered is geographic: whether the copying that matters occurred in Canada at all.
OpenAI’s position since 2025 is that its conduct, “the training of AI models and the automated crawling of web content,” happened entirely outside Ontario, and that Canadian copyright law does not reach activity outside Canada. Copyright is territorial: a Canadian court applies Canadian law to acts of copying or communicating that happen in Canada. If OpenAI persuades the court at trial that the reproductions made during training occurred only on servers outside the country, the jurisdiction win would not necessarily translate into liability.
The publishers argue the opposite: that their content is largely created in Ontario, is owned by Ontario-based companies, and that scraping happened in the province. Justice Kimmel found a sufficient connection to Ontario for six of the ten defendants and dismissed the claims against the other four. One law firm’s analysis summarized three presumptive connecting factors: that some OpenAI entities carried on business in Ontario by collecting data from Canadian websites and users and advertising to Ontario customers; that there was a good arguable case that torts including copyright infringement were committed in Ontario; and that the publishers’ terms of use were a contract made in Ontario that OpenAI allegedly breached.
Lawyer Barry Sookman’s analysis noted that the carrying-on-business finding leaned on the Equustek line of cases rather than decisions requiring physical presence, that the infringement analysis relied on a Supreme Court case about the communication right rather than the reproduction right, and that the contract finding did not examine how an autonomous crawler could accept browsewrap terms. The jurisdiction decision did not resolve the substantive issues, so those questions remain open on the merits.
Why Canadian Fair Dealing Offers OpenAI Narrower Cover Than US Fair Use
OpenAI has said its models rely on fair use, an open-ended test built on four fair use factors that can accommodate new kinds of uses. Canada’s fair dealing exception is narrower: a use must first fall within a listed purpose, such as research, private study, education, parody or satire, before a court asks whether the dealing was fair. Canada also has no text-and-data-mining exception of the kind the European Union has adopted.
Global News described the suit as the first case of its kind in Canada to address whether using copyrighted content to train generative AI is lawful, and the appeal ruling did not reach that question. OpenAI told the court that if U.S. courts find such training lawful under fair use, a contrary Ontario result under Canadian law would be problematic. Any defense of the copying would be assessed under Canadian fair dealing rather than U.S. fair use.
Can Canadian Courts Hear Copyright Cases Against Foreign AI Companies?
Increasingly, according to some lawyers. The Van Breda test asks whether a real and substantial connection links the dispute to Ontario, and the presumptive connecting factors include where a defendant operates a business, where the alleged wrong occurred and where a related contract was formed. Justice Kimmel relied on Equustek to find that some OpenAI entities carried on business in Ontario, a step Sookman says departs from decisions requiring a physical presence.
Miller Canfield lawyers Anita Marinelli and Robin Asher paired the OpenAI ruling with the Federal Court of Appeal’s 2024 decision in Munchkin v. Angelcare as showing a consistent approach by Canadian courts to asserting jurisdiction over U.S. companies.
The appellate ruling is relevant to other Ontario suits. Postmedia, Torstar, The Globe and Mail and The Canadian Press launched a copyright suit against Anthropic in an Ontario court in October; CBC/Radio-Canada is not part of that action. Toronto-based VerticalScope separately sued OpenAI in Ontario in May 2026 over the scraping of its online forums. Any defendant in those cases that weighs a jurisdiction challenge will now have to contend with an Ontario appellate ruling that went against OpenAI.
Claim Seeks Statutory Damages Across 10.65 Million Works
The coalition’s lawsuit was announced Friday, November 29, 2024, in the Ontario Superior Court of Justice, with seven plaintiffs in all: Toronto Star Newspapers Limited, Metroland Media Group, Postmedia Network, PNI Maritimes LP, The Globe and Mail, The Canadian Press and CBC/Radio-Canada.
The claim alleges copyright infringement, unjust enrichment and breach of contract based on the publishers’ terms of use. It seeks statutory damages of $20,000 per work, or an amount the court considers fair, along with OpenAI’s profits from the alleged infringement and an injunction. Twenty thousand dollars is the ceiling the Copyright Act sets for commercial infringements, with a floor of $500 per work. The publishers say they own and have published a combined 10.65 million works since 2015. The allegations are unproven, and no court has set a damages figure.
When the suit was filed, an OpenAI spokesperson said ChatGPT’s models are “trained on publicly available data” and that the company works with news publishers on how content is displayed, credited and linked, with an option for publishers to step out. Postmedia said OpenAI profits from the content without permission or compensation. OpenAI is represented by Gowling WLG, according to a Canadian Lawyer case listing, while Lenczner Slaght acts for the publishers.
Ottawa Backs Cohere, Itself Sued by Publishers, While Copyright Policy Stays Unwritten
The ruling arrives while federal copyright policy on AI remains unsettled. AI Minister Evan Solomon told The Logic in June 2025 that the government’s updated regulatory framework would address copyright, saying that in principle creators have a right to be compensated but that where the line sits is still being determined by the courts. When Prime Minister Mark Carney released the 50-page AI for All strategy on June 4, 2026, the word “copyright” did not appear, according to McCarthy Tétrault.
Parliamentary Secretary Taleeb Noormohamed called the strategy a living document and told Canadians to “stay tuned,” according to Global News, while Culture Minister Marc Miller said in March that he does not believe copyright law needs to be reopened for AI training. Law professor Michael Geist wrote that the strategy backs creators through a new $50 million Creative Technology Program rather than licensing mandates or limits on fair dealing. A House of Commons committee report in April 2026 recommended training-data transparency, opt-in consent for text and data mining, and licensing of copyrighted training material.
Industry groups say the gap is costly. News Media Canada chief executive Paul Deegan called the omission troubling and said it concerns music and book publishers as well as news publishers. The Coalition for the Diversity of Cultural Expressions, which says it represents more than 350,000 creators, argued government support should be conditional on generative AI companies complying with copyright law.
Ottawa’s industrial policy points the other way on the same technology. On October 29, Carney announced a commitment to invest up to C$1.6 billion in equity in Toronto-founded Cohere, to be made in stages with conditions covering Canadian jobs, computing infrastructure, headquarters and work with the Canadian AI Safety Institute. The government’s announcement does not mention copyright, even though a group of publishers that includes the Toronto Star, Condé Nast and McClatchy is suing Cohere in a U.S. court over the alleged collection of their articles without permission or payment. Australia took the opposite route in July, when Prime Minister Anthony Albanese committed to introducing AI copyright legislation in early 2027 and called training on creative work without the creator’s control over price and value theft, after Attorney-General Michelle Rowland ruled out a text-and-data-mining exception in October 2025.
US Rulings on News Training Could Diverge From Ontario Outcome
Copyright suits against OpenAI, including the one brought by The New York Times, are consolidated before Judge Sidney Stein in the Southern District of New York. On August 31, Stein ordered the Times to show cause why its case should not be stayed pending summary judgment motions in other cases in the consolidation. The Justice Department filed a statement of interest on September 1 that called training extraordinarily transformative, a non-binding position.
A U.S. win for OpenAI on fair use would not bind an Ontario court applying Canadian fair dealing, and an Ontario loss would not bind a U.S. judge. That asymmetry is the reason the venue fight mattered: the same training conduct could be treated differently under each regime.
Supreme Court Leave, Discovery and Legislation Remain Open Questions
OpenAI’s remaining route on venue is an application for leave to appeal to the Supreme Court of Canada, which grants leave where an issue is of public importance or otherwise warrants its decision. The application must be filed within sixty days after the judgment, with the month of July excluded from the count. The case has spent nearly two years on preliminary motions, including a partial sealing order OpenAI obtained in August 2025.
Discovery and any fair dealing arguments lie ahead. In the U.S. consolidation, Judge Stein in January affirmed an order requiring OpenAI to produce 20 million de-identified ChatGPT logs, and whether an Ontario court compels similar disclosure will shape how the publishers prove what was copied. A spokesperson for Miller said in June the government had no update on its plans for a text-and-data-mining statement; any move to add or rule out such an exception could reshape the merits of the Ontario suits mid-stream.
Frequently Asked QuestionsDid the Ontario Court of Appeal decide whether OpenAI broke Canadian copyright law?
No. The ruling decided only that the case can proceed in Ontario. Whether OpenAI infringed, and whether the copying happened in Canada, remain to be decided in later proceedings.
Does Canada have a fair use defense for AI training?
No. Canada has fair dealing, which covers listed purposes such as research, private study and education, and Canada has no text-and-data-mining exception. The OpenAI suit is described as the first Canadian case to address whether training generative AI on copyrighted content is lawful.
Can OpenAI still appeal the ruling?
OpenAI can ask the Supreme Court of Canada for leave to appeal, which is available where an issue is of public importance or otherwise warrants the Court’s decision. The application must be filed within 60 days of the judgment, excluding July.
How much money is at stake?
The publishers seek damages to be set at trial or statutory damages of up to $20,000 per work, and say they have collectively published 10.65 million works since 2015. No court has set a damages figure, and the allegations remain unproven.
Disclosure: The Canadian Press, whose wire report is cited in this article, is a plaintiff in the lawsuit, as is CBC/Radio-Canada. This article also relies on the independent Broadcast Dialogue report for the panel composition, the quoted finding and the costs award.