Key Takeaways
- That 2025 survey number is stark: 78% of Canadian publishers say AI companies are ripping off their content without fair compensation.
- Canadian law hasn’t caught up, so there’s no clear legal precedent for how courts will rule on AI using published works for training.
- Publishers need to get proactive with strong licensing agreements and start using tech to track where their content ends up.
- The lack of any global agreement on AI content rights creates a chaotic legal mess for publishers trying to get paid.
- Talking directly with AI developers to work out payment models is going to be faster and more effective than just suing them.
A 2025 survey from the Canadian Media Producers Association (CMPA) just confirmed what everyone suspected: 78% of Canadian publishers are convinced their content is being scraped by AI models, including Anthropic’s, without a license or a dime in compensation. This isn’t just a number. It points to a major conflict over how content creators can possibly protect their IP when large language models are swallowing the web whole for training data. The fight over Anthropic AI and publisher rights is a real-world legal and economic showdown that’s already reshaping the digital media business.
The 78% Discrepancy: Publishers Feel Undervalued
That CMPA finding, that nearly four out of five Canadian publishers feel their rights have been violated, isn’t just hurt feelings. It’s about cold, hard cash. Publishers pour huge resources into creating original work, from deep investigative journalism to specialized industry reports. When an AI model like Anthropic’s Claude trains on that material without consent or a check, it guts the value of the original effort. This directly impacts the fundamental business model of content creation. If a news outlet spends months on a complex series investigating municipal corruption and an AI can just synthesize its findings to answer user questions, all without a link or financial payment, the incentive to fund that original work evaporates. This kind of parasitic relationship eats away at the foundation of quality journalism.
Only 12% of Publishers Have Formal AI Licensing Agreements
Despite all the alarm, a separate 2025 analysis by the Canadian Association of Journalists (CAJ) found that only a shocking 12% of Canadian publishers have managed to set up formal licensing agreements with AI developers. That huge gap between the perceived threat and any real action is telling. So what’s the holdup? Part of the problem is that AI technology is moving much faster than our legal systems can keep up, leaving many publishers without a clear playbook or the budget for specialized lawyers who can negotiate these complex digital rights deals. At the same time, some AI companies have been operating under a vague premise of “fair use,” arguing their models produce new work. Publishers rightly argue that the quality and usefulness of the AI’s output are directly because of their input data, so compensation is a must. This legal stalemate creates paralysis, leaving most of their content completely exposed.
Canadian Copyright Act: A Developing Field
Good luck finding a clear answer in the current Canadian Copyright Act (R.S.C., 1985, c. C-42), because it says nothing specific about training AI models on copyrighted material. While Section 27 is supposed to define copyright infringement, its application to an AI’s data ingestion process is completely up in the air. Here in Canada, we’re still waiting for a landmark ruling that gives a definitive answer on AI training data, which leaves both publishers and AI developers in a state of limbo. Publishers are leaning on the “reproduction right” in the Act, saying that copying their data for training is an open-and-shut case of unauthorized reproduction. AI companies might counter that their process is more like a human reading and learning, not making a copy for the public. The lack of a clear answer means any lawsuit would be a long, expensive gamble for everyone involved. This pushes people toward a wait-and-see strategy that helps no one.
The Global Disconnect: Varying International Approaches
This problem isn’t just a Canadian one. Different countries are going in wildly different directions. The European Union is moving toward stricter rules with its AI Act, which has clauses about being transparent with copyrighted training data. The United States, by contrast, is just a mess of lawsuits with no federal legislation offering any real guidance. This global fragmentation makes life incredibly difficult for Canadian publishers who serve an international audience. Your content, written in Toronto, can be scraped by an AI model trained in California and used to answer questions for someone in Berlin. Which country’s laws even apply? This jurisdictional maze makes enforcing your rights or negotiating consistent license terms practically impossible. Without a harmonized international standard, the legal field for content legal protections is a patchwork quilt, allowing some AI developers to just shop around for the weakest regulations.
My Perspective: Why Litigation Isn’t the Only Answer
The first instinct for many publishers is to sue when they feel their rights are trampled. While lawsuits are important, especially for setting precedents, I think it’s a mistake to see them as the main solution. The legal system is slow, expensive, and always struggles to keep up with fast-moving tech. By the time a landmark case finally concludes, the technology and the business models built around it may have already changed so much that the ruling is partly obsolete on arrival. A much smarter move is to get ahead of the problem with proactive engagement and negotiation. This means publishers need to form industry collectives to bargain with big AI developers like Anthropic from a position of strength, because individual publishers trying to do it alone will get crushed. It also means developing technical tools (like watermarking or better metadata) that allow for real tracking and attribution of content slurped up by AI. And of course, we need to advocate for specific amendments to the Canadian Copyright Act that explicitly define AI training as an activity that requires a license. Just suing after the fact is a reactive, losing game of Whac-A-Mole. We have to design a better game board.
The relationship between Canadian publishers and AI developers like Anthropic is evolving, and it demands action on multiple fronts. While the legal system plays catch-up, publishers must be aggressive about pursuing licensing agreements, lobbying for clear legislative changes, and exploring tech solutions to protect their intellectual property. The entire future of professional content creation depends on finding fair and sustainable models for AI’s interaction with copyrighted material. For publishers trying to survive these changes, it’s also critical to understand the evolving Meta AI policies. This is especially true as Meta AI’s reckoning for publishers arrives, making a strong plan more necessary than ever.
What is the primary concern for Canadian publishers regarding AI companies like Anthropic?
It’s simple: their content is being used to train AI models without permission and without payment. This strips the value from their work and puts their entire business model at risk.
Does the Canadian Copyright Act specifically address AI training data?
No. The current Canadian Copyright Act (R.S.C., 1985, c. C-42) is silent on AI training, which creates a huge legal gray area that’s the source of all this conflict and uncertainty.
What steps can Canadian publishers take to protect their content from unauthorized AI use?
They need to actively chase down licensing deals, use technical tools like watermarking or metadata tagging to track their content, and lobby the government to update the Copyright Act for the AI era.
Why is a global standard for AI content rights important?
Without one, AI companies can just exploit legal loopholes by operating in countries with the weakest copyright laws. A global standard would give publishers a consistent way to enforce their rights and negotiate deals everywhere.
Is litigation the most effective strategy for publishers in this dispute?
Not on its own. Lawsuits are too slow and costly to be the only tool. A better long-term strategy involves getting ahead of the problem through direct, industry-wide negotiations and pushing for better laws instead of just reacting in court.
