AI companies have built their products using Canadian journalism without permission or compensation, and are now delivering that journalism to consumers as their own product, asserts an AI news audit undertaken by the Centre for Media, Technology and Democracy at McGill University.

The first large-scale empirical audit of how AI models use and distribute Canadian journalism, two studies were led by Taylor Owen, Beaverbrook Chair in Media, Ethics and Communication and Founding Director of the Centre, alongside Aengus Bridgman, Associate Director, Research and Director of the Centre’s Media Ecosystem Observatory.

Over February and early March, they tested four major AI models on 2,267 Canadian news stories in both English and French (18,134 queries in total) to measure what models have absorbed from their training data and whether they attribute it. They also enabled web search, asking the same models about 140 specific recent articles from seven Canadian outlets across 3,360 experimental conditions, to measure whether AI models produce viable substitutes for current journalism and whether they credit the source.

The audit found that when asked about Canadian news events drawn from their training data, ChatGPT, Gemini, Claude, and Grok provide no source attribution 82% of the time. When given web access and asked about specific recent articles, the same models covered enough of the original reporting to substitute for the source in 54 to 81% of cases. Models linked to Canadian news sites in 29 to 69% of responses, but named the originating outlet  in only one to 16% of cases.

Owen and Bridgman call the pattern “a structural shift in how the informational value of journalism is captured and distributed.”

“AI companies have built commercial products that depend, in significant part, on the reporting that Canadian journalists produce,” the report states. “They have done so without compensation, without attribution, and without any obligation to sustain the infrastructure they are drawing from. The result is a system that accelerates the economic decline of the journalism it relies on.”

AI companies ‘absorbing the substance of journalism’

The audit’s authors say unlike issues that arose with the rise of social media that saw platforms capture advertising revenue by aggregating attention around news content, AI companies are “absorbing the substance of journalism, and delivering it directly to consumers as their own product.”

“The consumer’s need to visit the source is not just reduced by algorithmic demotion, as it was with social media. It is rendered unnecessary by the AI’s response itself,” the audit states. “The democratic implications extend beyond the economics of newsrooms. When AI models deliver confident answers about Canadian public affairs that are drawn from journalism but stripped of source and context, the public’s ability to assess the reliability of the information it receives is diminished. Canadians are increasingly receiving information about their own politics and society through (largely American) systems that neither credit the source nor guarantee accuracy.”

The audit says existing copyright and media policy frameworks are not designed to address this. While the Online News Act established the principle that tech companies profiting from the work of Canadian journalists should enter into a fair process to determine the value of the exchange, the audit’s authors say “there is no obvious equivalent metric for what AI companies do when they absorb reporting into model weights and deliver derivative answers to consumers.”

“The $100 million annual Google agreement under C-18 provides a reference point for the scale of compensation at stake, but the mechanism for arriving at obligations for AI companies is genuinely unresolved. The Copyright Act provides the other potential lever, but its application to AI training remains entirely unsettled. The Act’s fair dealing doctrine is limited to enumerated purposes, and whether large-scale commercial AI training constitutes ‘research’ has never been tested in Canadian courts,” the report states.

Its authors write that while a publishers’ lawsuit against OpenAI is cleared to proceed in Ontario, statutory licensing – which is already being investigated by the World Intellectual Property Organization – may be a better path forward than the courts, potentially setting rates through a centralized mechanism rather than deal by deal.

“None of these questions have easy answers, and the trade-offs involved are genuine. But the cost of deferring them is also real. In the absence of deliberate policy choices, the terms of AI companies’ relationship to Canadian journalism are being set by corporate design decisions made outside Canadian jurisdiction. The evidence we present here makes the scale of that relationship visible. What democratic institutions do with that evidence is a political choice, not a technical one.”