Home / Publications / Blog / ForIntel Research / Canada's biggest newsrooms sue Anthropic over Claude's training data

Canada's biggest newsrooms sue Anthropic over Claude's training data

Postmedia, Torstar, The Globe and Mail, and The Canadian Press say Anthropic scraped their journalism to train Claude without a license.

By Foragentis teamPublished 2026-10-211 min read

Four of Canada's largest news organizations sued Anthropic in an Ontario court on October 6, 2026. Postmedia, Torstar, The Globe and Mail, and The Canadian Press say Anthropic used their journalism to develop its Claude models without a license, and that it keeps doing so.

The statement of claim describes Anthropic as a company that "deliberately scrapes (i.e. accesses and copies) enormous amounts of text data" from the publishers' websites, tied to user prompts, "all without obtaining a valid licence from any of the News Media Companies." The claim also alleges Anthropic breached and circumvented technological measures the publishers had put in place to stop people from copying their content. It says the company "has done nothing to curtail its unlawful behaviour." Anthropic did not respond to requests for comment, and no court has ruled on the allegations yet.

This is the second time this exact group has gone to court over the same question. The same four publishers, together with CBC/Radio-Canada, sued OpenAI in 2024 over similar allegations. That case remains unsettled. Two lawsuits, two years apart, against two different companies, over one unresolved question in Canadian law: whether training a commercial AI model on a newsroom's journalism needs that newsroom's permission.

The lawsuit is about what a model is allowed to learn from without asking. ForIntel runs on Anthropic's models, among others, which puts Foragentis on the other side of the same question these publishers are asking in court: who gets credit for the words a model produces. We've already settled our part of that question: every claim in a ForIntel report traces back to a source the reader can click and check. No court, in Canada or anywhere else, has settled what a model is allowed to learn from in the first place, and this case will not settle it either. It adds one more data point to a fight still being fought in court, case by case, model by model.

A German Court Held Google Liable for a False AI Answer About a Real Doctor

A German court ordered Google to stop an AI Mode answer that wrongly named a doctor a convicted sex offender, after a similar May ruling against AI Overviews.

Publishers Take AI's Stealth Bots to Congress

Over 300 publishing executives, including Conde Nast and Hearst, asked Congress on September 29-30, 2026 to pass the Stealth Bot Prohibition Act. Here is what the bill does and does not do.

A chatbot hallucinated nuclear weapons cargo on a Chinese ship, and troops nearly boarded it (CNN exclusive)

CNN's exclusive reporting: during the U.S.-Iran war, an AI chatbot hallucinated nuclear cargo on a Chinese vessel, and the fabricated report nearly triggered a military boarding. The failure traces to one missing engineering requirement.

An Appeals Court Upheld a Copyright Win Over AI Training. The AI Company Lost.

A US appeals court upheld Thomson Reuters' copyright win over Ross Intelligence on September 29, 2026, reported as the first appellate decision on AI training and fair use. It covered 2,243 Westlaw headnotes rather than the whole case. Here is what the court decided, what it left open, and what to do about it.