OTTAWA — Quebec correspondent “Geneviève Tremblay” — who has been reporting on multiple news stories a day lately — seemed to be experiencing a glitch.
She wasn’t the only one. All of We News’ reporters and editors seemed to be malfunctioning on Friday, sending out error messages whenever The Canadian Press tried to ask them questions via chat.
That can happen when your “journalists” are — like the virtual staffers at We News — just fake personas generated by artificial intelligence.
AI has infiltrated news media in Canada and around the world. In early September, Nunatsiaq News published a story about an AI persona posing as an Inuk journalist in Nunavut — one of many personas used to repurpose work from legitimate news sources, including The Canadian Press, for We News.
The site describes itself as a “round-the-clock news portal covering Canada and the world.” But while it remains online, it appears to have been delisted from Google’s search results.
In late September, the story of “Daniel Robson” emerged. Canadian news outlets pulled his work from their websites after being told Robson appeared not to be a real person.
Policy Options editor-in-chief Les Perreaux was the first to sound the alarm in Canada. In an online post on Sept. 24, Perreaux wrote that an editor at Türkiye Today contacted him to ask him to vouch for Robson — and he realized he couldn’t.
“The key moment where my heart sank was when the Turkish journalist told me that Daniel Robson refused to do a video call,” Perreaux said in an interview.
Perreaux said he reviewed the correspondence and material Robson sent and spotted signs of AI product. He said the bigger issue was the attempt to deceive news editors and readers, while AI merely acted as an “accelerant.”
“AIis definitely part of the story, but it’s a smaller part than the deception part where there was a human behind this, pretending to be someone he’s not, or at least refusing to be who he really is,” he said.
In a story published on Sept. 29, Türkiye Today said the purpose of the Robson persona seems to be to attack Moroccan dissidents.
We News isn’t hiding its use of AI and clearly labels its “journalist” profiles — including photos of what appear to be human beings — as AI generated.
The AI personas at We News, which include grey-haired editor-in-chief “Gordon Halloway,” repurpose international, national, provincial and local news in more than 100 cities across Canada.
The Canadian Press did attempt to reach a human behind the site, but the contact email address listed on the site bounced back.
When the AI detection company Originality.ai analyzed just over 200 articles by “Geneviève Tremblay,” it identified them as AI-generated.
Originality.ai CEO Jon Gillham said the articles were rewritten versions of stories published by legitimate news media outlets like CBC, CTV, CityNews, Global and the Montreal Gazette. The Canadian Press also found its own reporting among the sources.
“I think industrial-scale AI slop (sites) like this one run the risk of degrading the trust in the news,” Gillham said.
The phenomenon isn’t limited to Canada. On Sept. 17, the U.S.-based website Futurism published a story about Brown Brothers Media, a company Futurism said bought up legitimate news sites and flooded them with AI-generated content — and claimed to pull 50 million page views a month.
Ahmed Al-Rawi, a professor of news, social media and public communication at Simon Fraser University, said sites with fake AI journalists are trying to monetize content.
“A lot of players, whether they are bad players or just profiteers, people who are really trying to get money, are exploiting the hard work of real journalists in order to repackage news content,” he said.
Al-Rawi said social media platforms “should also be blamed for enhancing the popularity of these fake websites.”
Western University assistant professor of journalism Chris Arsenault said it’s difficult to regulate these websites.
“You can get a free or very cheap AI model to pump out this kind of nonsense, and you can always find somebody to host it,” he said. “So it’s a very hard thing to legislate around or to respond to in terms of effective policy.”
Paul Deegan, president of the publishers group News Media Canada, said while news content is covered by copyright, it’s very difficult for publishers to sue.
“It’s expensive, it takes a long time, whether or not they are in fact in Canada,” he said. (We News says it’s based in Hong Kong.)
News Media Canada is calling on the federal government to state publicly that it does not intend to introduce a broad exception to copyright law for text and data mining.
AI Minister Evan Solomon’s office said in a media statement that “Canadians deserve to know when the information they consume has been generated or altered by AI.”
“Generative AI can be a useful and powerful tool, but Canadians should not be deceived by content that impersonates journalists or conceals how it was created,” the statement said.
Solomon’s office pointed to a recently-concluded consultation on AI transparency and said it recognizes “the concerns raised by journalists and other creators about consent, compensation, attribution and transparency when their work is used to train AI systems.”
There may be no silver bullet to solve the problem, but Arsenault said there is one antidote: old-fashioned reporting.
He noted the “Daniel Robson” case came to light because “of an old-school editor in Turkey who had the temerity and good sense to just, God forbid, pick up a telephone and do some basic journalism, just do some due diligence on a would-be contributor.”
Futurism reported on Sept. 30 that the Brown Brothers Media sites now appear to have stopped publishing and were delisted or significantly downranked by Google.
In Canada, We News also appears to have been delisted. When The Canadian Press searched for the website using Google’s search engine, it got no results — but the site immediately popped up as the top result on another search engine.
Google did not directly answer questions about whether it has delisted We News from its search results, when it may have done so, and whether that was the result of news coverage about its AI “journalists.”
A spokesperson said in a statement the company’s policies for search and Google News “prohibit producing low-value content at scale to game our ranking systems.”
“While we generally don’t comment on individual sites, we act on policy violations to keep 99 per cent of search visits spam-free,” the spokesperson added.
Deegan said it’s in Google’s interest to “make sure that when people go to Google for information that they’re getting accurate information … For them to de-rank and remove these sites that are clearly involved in theft makes sense.”
But Arsenault questioned whether Big Tech companies should be “playing hall monitor and being arbiters of truth.”
“They have a role to play in taking down stuff that’s clearly untrue … But I don’t know if we want them making those editorial judgments.”