WE NEWS Invented Indigenous "AI Correspondents" to Rewrite Real Northern Reporting

A website called WE NEWS has built a roster of 121 "AI correspondents," and at least three of them are dressed up as northern Indigenous journalists. According to CBC News, the profiles for Kaylee Johns, Warren Lafferty and Leah Qammaniq claim Yukon First Nation, Dene and Inuk backgrounds, yet all three are generated by artificial intelligence. The articles under their names are largely reworded versions of reporting by real newsrooms, illustrated with AI images.

This is not a story about a new model or a clever product. It is a story about what happens when cheap text generation meets the old content-farm playbook, and when the farm borrows a community's identity to look credible. It also sits next to a wave of similar problems we have covered, from AI slop flooding bug-bounty programs to the arXiv submission cap aimed at the AI paper flood.

TL;DR: what is confirmed and what is not

The facts above come from CBC's reporting and from Canadian HR Reporter's summary of the Nunatsiaq News original. We did not independently test the site, and we are not linking to it.

What exactly did CBC report?

CBC reporter Francis Tessier-Burns found that WE NEWS lists three northern Indigenous "AI correspondents." Their profiles say they are "a member of Yukon First Nation," "of Dene heritage," and "an Inuk raised in the North." The site's About Us page says it "relies on artificial intelligence technologies" and uses "the assistance of AI editorial agents, operating under clear editorial principles: fact verification, transparency about sources, independence of the editorial line, and prompt correction of errors."

In practice, CBC says, the process seems to be repurposing articles from legitimate news sources, rewording them and publishing the edited version with AI-generated images. CBC also found at least one error that shows the missing verification step: an article about a shooting that purportedly happened in the Yukon, whose sources actually concerned the city of Yukon in Oklahoma.

That last detail is the practical tell. A human regional reporter would not confuse a Canadian territory with an Oklahoma suburb. An automated rewriter fed a headline with the word "Yukon" can.

Why the Indigenous identity matters

The cheapest content farms use generic bylines. This one chose identities that carry meaning. Kerry Benjoe, interim president of the Indigenous Media Association of Canada, told CBC the profiles disturbed her. Indigenous people were historically excluded from media, she said, and generated personas take over "the space [Indigenous journalists] fought so hard to get." She also described it as "stealing authority," because the personas compete for the same audience that real Indigenous outlets fight to earn.

Tamara Voudrach, executive director of the Inuvialuit Communications Society, argued that even factually correct articles are not authentic, because a real Indigenous journalist carries the values they were raised with into editorial decisions. Her organization has operated since 1983, and she called an AI entity that passes itself off as Indigenous "just another form of colonialism" and "erasure."

These are the views of the people quoted, not findings of fact. But they explain why this case drew attention beyond ordinary scraping complaints: the fake persona is the product. A face and a heritage line turn a rewritten wire story into something that reads as lived local expertise.

How the model works: rewrite, relabel, monetize

Strip away the personas and the architecture is simple, which is why it scales to 121 bylines.

1. Ingest. Monitor feeds or sitemaps of real outlets for new articles.
2. Rewrite. Ask a language model to paraphrase each story so it no longer matches word for word.
3. Dress. Assign it to a persona whose invented background fits the beat, and generate a headshot and article images.
4. Publish and monetize. Run the result with ads, and capture search traffic that would have gone to the original.

Ahmed Al-Rawi, director of the Disinformation Project at Simon Fraser University, told CBC that the disclosure at the bottom of each article is not enough. "A lot of people might read the article … and think that it's real," he said, adding that the lack of clear attribution to the original authors is a second problem, since the site depends on the work of real journalists to generate revenue.

Corey Larocque, managing editor of Nunatsiaq News, whose stories have been mined heavily, said it is hard enough for news organizations to make a go of it without an outside AI source using their content and conceivably taking away advertising dollars.

Is this legal? The honest answer is "unclear and slow"

Paul Deegan, president of News Media Canada, told Nunatsiaq News that "the theft is blatant and it's happening on an industrial scale," according to Canadian HR Reporter. His practical point is the important one: the only current remedy is a lawsuit, and that is neither affordable for most publishers nor simple when the accused operates across several borders, here Hong Kong registration, French hosting and an unreachable publisher.

Brent Jolly, president of the Canadian Association of Journalists, called the fight "a game of whack-a-mole" and asked Ottawa to strengthen copyright law so that repurposing news content is explicitly illegal. Canadian HR Reporter adds that Jolly and Deegan were disappointed that copyright protection for news was absent from Canada's new AI strategy, released earlier this summer, which commits more than two billion dollars to expand AI capacity.

Two caveats are worth stating plainly. First, rewording is not automatically infringement; copyright protects expression, not facts, and courts are still sorting out where machine paraphrase falls. Our AI lawsuit tracker shows how unsettled that question is even for large, well-funded defendants in the United States. Second, no regulator or court has ruled on WE NEWS. Everything legal here is argument, not outcome.

What readers and publishers can actually do

For readers, a few quick checks cut through most AI-written sites:

• Find the masthead. Real outlets name editors and a physical address you can verify somewhere else.
• Reverse-search the author photo. Generated faces have no history; real reporters appear on social profiles, conference pages and other publications.
• Look for original reporting. If every article is a paraphrase of a story you can find at a named outlet, go to the outlet.
• Check ownership. CBC traced the legal notice page; a domain lookup is a few seconds of work.
• Read the footer, but do not stop there. A disclosure that the text was written with AI does not tell you whether the author is real, whether the facts were checked, or whose reporting was reused.

For publishers, the options are limited but real: document copying with timestamps, file takedown notices with the host and ad networks (the money trail is often the fastest lever), and publish canonical links and bylines that make the original authorship easy to prove. Industry bodies like News Media Canada are asking for stronger law; that is a slower path.

Identity verification card with a check mark, illustrating how to confirm that a journalist byline belongs to a real person

The bigger pattern: synthetic people as a feature

WE NEWS is one example of a wider trend. Synthetic identities are being deployed for authority rather than for output quality. We covered a similar dynamic when China began piloting ID cards for AI digital humans, where the question was how a society labels non-human personas at all. Influence operations use the same trick: OpenAI described false-front accounts built for Russia- and Iran-linked campaigns, with personas designed to look like ordinary local voices.

There is also a quieter counter-trend. Audiences are starting to pay a premium for things made by identifiable people, which we explored in the BBC and Deezer discussion of human-made art. And inside the publishing industry itself, staff have objected to undisclosed AI use, as in Wired's report on the Big Five publishers using AI for covers and copy. Disclosure norms are being fought over everywhere, and WE NEWS shows what a minimal, technically compliant disclosure looks like when the surrounding design is built to mislead.

Where AI-agent security fits

This case is content deception, not an agent breach, but the control problem is related: systems that can publish at scale need guardrails before they act. If you build or deploy agents that write or post on your behalf, AgentBeam, the agent security platform from the explainx.ai team, is designed to stop AI agents before they take dangerous actions, which includes publishing actions you did not intend.

What to watch next

• A response from AZ Medias or the named publisher. None has been reported.
• Platform and ad-network action. Removing monetization is usually faster than a court case.
• Copyright policy in Canada. Industry groups want news content addressed explicitly in AI policy.
• Copycats. Jolly's whack-a-mole comparison is the realistic baseline: one site going dark does not stop the next.

The takeaway for anyone building with language models is uncomfortable but simple. The same tooling that powers a helpful research assistant can produce a believable fake newsroom in an afternoon, and the harm lands on communities that spent decades earning the right to tell their own stories.

Related reading

• Google OSS VRP freezes bug bounties over AI slop reports
• arXiv caps submissions as the AI paper flood grows
• China pilots ID cards for AI digital humans: claimed vs verified
• OpenAI disrupts Russia and Iran false-front influence operations
• AI lawsuit tracker: 248 cases against 76 companies
• People pay a premium for art made by humans
• Big Five publishers quietly using AI for covers and copy

Details reflect CBC and Nunatsiaq News reporting as of September 2026 and may change; check the linked sources for updates.