Canada's four largest news publishers have filed a copyright lawsuit against Anthropic in Ontario Superior Court, accusing the AI company of scraping roughly 13.5 million of their articles to train and run Claude without paying a cent. The case, filed on October 6 and first reported on October 19, 2026 at 12:47 (UTC-4), names Postmedia, Torstar, The Globe and Mail, and The Canadian Press as plaintiffs.
The statement of claim paints a familiar picture — an AI company hoovering up journalism at industrial scale — but the jurisdiction makes it different. Canada has no American-style "fair use" doctrine. Its narrower "fair dealing" defence is far harder to invoke for commercial model training, and a US appeals court has already ruled against one AI firm that trained on a publisher's content. That combination could make this the most dangerous copyright case Anthropic has faced yet.
What the publishers say happened
The four companies collectively produce a large share of Canada's English-language journalism, from The Globe and Mail's national broadsheet to Postmedia's chain of dailies and The Canadian Press's wire service. Their claim says Anthropic "deliberately scrapes enormous amounts of text data" from their websites, ingests it into Claude, and then keeps scraping on an ongoing basis to answer user prompts.
They make four legal claims:
- Copyright infringement — copying journalism to train and operate Claude under Canada's Copyright Act.
- Breach of contract — ClaudeBot and other crawlers ignored website terms of service that limit content to personal use.
- Circumventing protections — bypassing paywalls, logins, and other technological measures designed to stop copying.
- Unjust enrichment — Anthropic built a multibillion-dollar business on their work without compensation.
The scale is staggering. By the publishers' count, about 13.5 million works have been taken since 2021: roughly 4.2 million from The Canadian Press, 3.5 million each from Postmedia and Torstar, and 2.3 million from The Globe and Mail. They're asking for the statutory maximum of C$20,000 per infringed work — a figure that, applied across the full count, runs into the billions — plus punitive damages and a permanent injunction barring Anthropic from using any of their current or future content without written consent.
"Rather than seek to obtain the content legally, Anthropic has elected to brazenly misappropriate the News Media Companies' valuable intellectual property," the claim reads, according to Cartt.ca, which first obtained the statement of claim.
Why this case is different from the others
This isn't the first AI copyright lawsuit, and it isn't even the first one from this group. The same publishers — joined by CBC/Radio-Canada — sued OpenAI in Ontario in November 2024, and that case remains unsettled. But the Anthropic case carries sharper teeth for three reasons.
Canada's legal framework is less forgiving. In the US, AI companies have leaned on fair use to argue that training copies are transformative. Canada's fair dealing is a defined list of purposes (research, private study, criticism, review, news reporting, education, parody, satire) — commercial AI training doesn't obviously fit any of them. A ruling that model training requires a licence in Canada would force every AI company to renegotiate how they access Canadian content, and could embolden publishers in other jurisdictions with similar regimes.
Anthropic's own track record weakens its position. The company agreed last year to pay $1.5 billion to settle a US class action from authors over millions of books downloaded from pirate sites. A judge ruled that training on legally purchased books was fair use — but keeping pirated copies was not. Anthropic has also publicly pursued licensing deals with some publishers in other countries, which undercuts any argument that licensing is impossible or unnecessary. The claim pointedly notes that "other AI companies have engaged meaningfully with publishers in other jurisdictions, resulting in legitimate market-based licensing solutions."
The timing is brutal. Anthropic is weeks away from a potential IPO that could value the company at $1.5 trillion or more. It just signed its first Canadian data centre in Nova Scotia, deepening its physical presence in the country. And in Australia, Anthropic asked the government for a copyright exception to train AI — a request the public broadcaster ABC said would "cannibalise" news. A Canadian lawsuit alleging the same behaviour the company is asking permission for elsewhere is the kind of narrative plaintiffs' lawyers dream about.
The competitive context
| Company | News publisher lawsuits | Key settlement | Licensing deals |
|---|---|---|---|
| OpenAI | NYT, multiple US suits, Canadian (2024) | — | Axel Springer, News Corp, others |
| Anthropic | Canadian (2026) | $1.5B US authors class action | Some, fewer than OpenAI |
| Multiple US suits | — | Various publishers | |
| Meta | Multiple US suits | — | Fewer public deals |
Sources: Company announcements, Reuters, The Canadian Press
The pattern is clear: OpenAI moved early to sign licensing deals with major publishers, while Anthropic has been more selective. That strategy saved money in the short term but leaves the company more exposed now that publishers are organised and willing to litigate. The Canadian suit could force Anthropic into the same licensing posture — or into a long, expensive court fight that runs parallel to its IPO roadshow.
What to watch
Anthropic has not yet filed a defence and did not respond to requests for comment. The allegations are unproven. But the next several weeks are consequential:
- Will Anthropic settle quickly? A pre-IPO settlement would remove a headline risk, but the $1.5B authors' settlement set a high floor.
- Will the Ontario court grant an injunction? A permanent ban on using the publishers' content would force Claude to retrain or filter — a costly proposition either way.
- Does this spread? If Canadian publishers win even a preliminary ruling, expect similar suits in the UK, Australia, and Brazil — all jurisdictions without broad fair use.
- What does it mean for the OpenAI case? The same plaintiffs are already suing OpenAI in the same court. A ruling in either case could set precedent that binds the other.
The core question isn't whether AI companies should be able to use journalism. It's whether they should pay for it. Canada's courts may be the first to give a definitive answer — and the answer could cost Anthropic billions.
Sources: The Canadian Press via CTV News (Oct 19, 2026), Cartt.ca (statement of claim, Ontario Superior Court of Justice, Oct 6, 2026), MadRobot.blog analysis (Oct 19, 2026), Broadcast Dialogue (works counts and relief sought)
No comments yet