Canada's largest newspapers sue Anthropic in Ontario over Claude training

Who is suing and what they claim

A group of Canadian news publishers that includes the country's largest newspapers is suing Anthropic in an Ontario court, alleging the company copied their journalism without permission to build and run its Claude models. The plaintiffs are Postmedia, Torstar, The Globe and Mail and The Canadian Press, according to a Canadian Press report published by Global News on 19 October 2026.

The statement of claim was filed on 6 October and was first reported by the Canadian trade site Cartt.ca on 13 October. Anthropic had not responded to the Canadian Press's request for comment, and none of the allegations has been tested in court.

The claim describes the plaintiffs as responsible for "a significant portion of Canada's journalistic content," from daily print broadsheets to round-the-clock digital news, with millions of readers. Its central complaint is commercial: "Anthropic has capitalized on the commercial success of its AI models, building an expansive suite of services, and raised significant capital — all without obtaining a valid licence from any of the News Media Companies."

The legal theories in the claim

As summarised by the Canadian Press, the claim makes four kinds of allegation. The first is copyright infringement: that Anthropic copied the publishers' articles and used them, without authorisation or consent, to develop its models. The second is breach of contract, on the basis that scraping the sites violated their terms of service.

The third is that Anthropic circumvented technological protection measures the publishers put in place to restrict copying. That matters in Canada because section 41.1 of the Copyright Act separately prohibits circumventing such measures, and gives copyright owners the same remedies, including injunctions and damages, that are available for infringement. The claim, as reported, does not spell out which measures were bypassed or how.

The fourth is unjust enrichment. "In doing so, Anthropic has been substantially and unjustly enriched to the detriment of the News Media Companies," the claim says. The public reports do not state what damages or other relief the publishers are seeking.

Training data and live retrieval

The claim targets two different ways Claude can touch a news site. One is training: "To obtain content for ingestion into its models and services, Anthropic deliberately 'scrapes' (i.e. accesses and copies) enormous amounts of text data from the internet and from the Plaintiffs' Websites," it says.

The other is live retrieval. The publishers allege Anthropic also scrapes content on an ongoing basis in response to user prompts. Anthropic's own documentation on its web crawlers describes three separate robots: ClaudeBot, which collects web content that could contribute to model training; Claude-User, which fetches pages when a person asks Claude a question; and Claude-SearchBot, which indexes content to improve search results. Anthropic says all three honour robots.txt.

That split matters legally and commercially. A training copy is made once and absorbed into a model. A user-triggered fetch can pull today's article into an answer within seconds, which competes more directly with a visit to the publisher's site. The claim also argues that Anthropic "has done nothing to curtail its unlawful behaviour" despite other lawsuits and public criticism.

What the publishers' robots.txt files show

AI Lookout checked the robots.txt files of three plaintiffs on 19 October 2026. The Globe and Mail's file disallows ClaudeBot, Claude-User, Claude-SearchBot, Claude-Web and anthropic-ai across the whole site. The Toronto Star's file, for Torstar's flagship, blocks the same five agents. The Canadian Press site blocks ClaudeBot, Claude-User, Claude-Web and anthropic-ai.

The National Post's robots.txt, for Postmedia's national paper, did not name any Anthropic agent when we checked, though it blocks several other AI crawlers. A robots.txt file shows only the current rules, not when each block was added, so these checks cannot say whether Anthropic's bots were blocked when the alleged copying happened. Robots.txt is also a voluntary convention rather than a technical lock, which is part of why the claim's separate allegation about protection measures will be closely watched.

The same coalition is already suing OpenAI

The same four publishers, together with CBC/Radio-Canada, sued OpenAI in 2024 in a similar case. The Canadian Press reports that case has yet to be settled. The new claim leaves out the public broadcaster but otherwise extends the coalition's strategy from the maker of ChatGPT to the maker of Claude.

The publishers have also been pressing Ottawa to take a position on how AI systems use copyrighted work. According to the Canadian Press, the federal government's national AI strategy did not address the issue, which leaves the courts as the main venue for now.

Anthropic has been expanding its physical presence in the country, including its first Canadian data centre lease in Nova Scotia, so the case lands as the company deepens its commercial ties to the market where it is being sued.

Anthropic's copyright history in the US

In the United States, Anthropic has already been through a major copyright fight, over books rather than news. In Bartz v. Anthropic, an authors' class action in federal court in California, the court granted final approval of a settlement on 20 July 2026, according to the official settlement website. The administrator says it expects to send a first round of payments on or before 15 November 2026.

That case does not decide anything in Canada. Canadian copyright law uses a fair dealing test with specific listed purposes rather than the broader US fair use doctrine, and the Ontario claim adds contract and protection-measure allegations that stand somewhat apart from the training question. But it does show that Anthropic has been willing to pay to resolve large copyright claims.

The company is facing scrutiny on other fronts too, including the FTC's demands to Anthropic and OpenAI over AI safety incidents. Publishers worldwide are also renegotiating how platforms crawl their sites, as seen in Meta's rewrite of its business crawling terms.

What happens next and what is unknown

Anthropic has not yet filed a defence, and the public reports do not say whether it will challenge the Ontario court's jurisdiction, which is a common early step for foreign companies sued in Canada. The full statement of claim has not been published online, so AI Lookout's account relies on the passages quoted by the Canadian Press.

Important details are missing: the remedies sought, the time period covered, which Claude models are named, and what evidence the publishers have that their protection measures were bypassed. Those points will shape how much the case is worth and whether it focuses on past training or on the live fetching that continues today.

Why it matters

The case puts Claude in the same legal frame as ChatGPT in Canada, brought by the same publishers. If the court accepts the argument that user-triggered fetching is a fresh act of copying each time, every AI assistant that browses the web for answers could face similar claims, regardless of how its training data was gathered.

For publishers, the case is leverage in a market where licensing deals are becoming normal. For AI companies, it is a reminder that respecting robots.txt going forward may not settle claims about what was collected before, and that contract and anti-circumvention arguments can travel alongside copyright.