AI systems use Canadian journalism but don't attribute sources: report

The ChatGPT app icon appears on a smartphone screen on Monday, Aug. 4, 2025, in Chicago. (AP Photo/Kiichiro Sato)

OTTAWA — A new study released on Monday says AI systems depend on Canadian journalism for the information they provide users but don’t offer compensation or proper attribution in return

Researchers at McGill University’s Centre for Media, Technology and Democracy tested 2,267 Canadian news stories on major AI models.

“All four models showed extensive knowledge of Canadian current events consistent with having ingested Canadian news reporting,” the report says.

The researchers found when ChatGPT, Gemini, Claude and Grok were asked about Canadian news events from their training data, they did not provide source attribution about 82 per cent of the time.

They said when the AI platforms were asked about specific articles with web access enabled, in most cases they provided enough of the original reporting in their responses to make it unnecessary for the user to visit the original news source.

While half of the responses included at least one Canadian link, they only named a Canadian source 28 per cent of the time, the report said.

“Links provide a pathway back to the source, but consumers reading the response itself rarely see an indication of whose journalism they are consuming,” the report said.

The report said AI companies now extract value from journalism “at every stage: ingesting news archives as training data, producing derivative content without naming the sources, and delivering answers to consumers that could reduce the need and incentive to visit the original source.”

The system “accelerates the economic decline of the journalism it relies on,” the researchers said in an accompanying policy brief.

The report was released during a national summit the federal government is holding in Banff to look at issues involving AI and culture.

In opening remarks at the summit, Culture Minister Mark Miller said there are “legitimate questions” around copyright, market-based licensing and AI-generated content flooding the marketplace.

Artificial Intelligence Minister Evan Solomon said during the government’s recent consultation on its upcoming AI strategy, it heard “creators need assurance that AI will be developed with guardrails.”

He added that there are “real questions” about such issues as copyright, ownership and data mining.

Solomon said he and Miller are “open to these conversations.”

A coalition of Canadian news outlets, which includes The Canadian Press, Torstar, The Globe and Mail, Postmedia and CBC/Radio-Canada, are suing OpenAI in an Ontario court.

They argue OpenAI is using their news content to train ChatGPT, breaching copyright and profiting from the use of that content without permission or compensation.

In 2023, Ottawa passed the Online News Act, which requires Meta and Google to compensate media outlets for displaying their content. Meta pulled news off its platforms in response, while Google has started making payments under the act.

The researchers’ policy brief says the problems posed for journalism by social media and AI systems are distinct.

While social media platforms “captured advertising revenue by aggregating attention around news content,” it reads, “AI companies are doing something different: they 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.”

Anja Karadeglija, The Canadian Press