Images of Canada's most wanted are AI-generated. How do we know they're accurate?
The lifelike images blink, turn their heads, cross their arms, squint and shrug.
To some, the new reveal of Canada's top 25 most wanted fugitives might be reminiscent of the Harry Potter posters warning that Sirius Black escaped from Azkaban. But while the moving images of the men and women used by the Bolo Program aren't magical, they also aren't real — they're AI.
"This year, with the help of AI, we bring our top 25 fugitives into this room, life-sized, moving as they might if you saw them on the street, at the supermarket, at [the] airport or anywhere else," Bolo Program executive director Max Langlois said in a Toronto news conference Thursday.
Behind him, AI-generated images of 25 people wanted for some of Canada's most serious crimes stood in sweatshirts and jeans, put their arms behind their backs, shuffled and sighed.
"If it feels uncomfortable, that's the point," Langlois added.
But some security and technology experts caution that the use of AI to present images of Canada's most wanted fugitives isn't without risks — such as the accuracy of the digital avatars, privacy concerns and whether there are enough safeguards around such third-party programs working with police.
The Bolo Program, which stands for "be on the lookout," has been partnering with police forces in some of Canada's biggest cities since 2018. The national initiative uses public awareness campaigns, new technologies and cash rewards to amplify police requests for the public's help in locating major crime suspects.
The program also uses gimmicks to encourage public engagement. Last year, for instance, police wore "spoof masks" bearing the faces of the 25 most wanted in a Bolo news conference. Bolo has also utilized life-size cutouts, online contests, advent calendars and billboards.

Police officers wearing masks bearing the faces of wanted fugitives stand during a Bolo Program news conference, in Vancouver, on Oct. 8, 2025. (Darryl Dyck/The Canadian Press)
Bolo has been using AI-generated images for a few years now, Langlois told CBC News in an email statement, explaining that they use the available pictures and descriptions of the suspects from police to ensure their accuracy.
The AI images give the public a "better idea" of how the suspects may look, he said.
Experts question accuracy of avatars
While this specific use of AI could potentially be helpful, we also don't have all the information about what data was used to generate these avatars, said David Eliot, a PhD candidate in criminology at the University of Ottawa who researches AI, citizen rights and security.
When asked how the AI images were generated, Langlois said that information was confidential.
Eliot questions how Bolo verified the wanted individuals' mannerisms and body types — how much is assembled from collected data or photos, how much from witness descriptions and how much from AI or human assumptions.
How forensic sketch artists are adapting technology to bring their subjects to life
A sketch to remember: RCMP forensic imaging expert says drawing perpetrators is emotional work
"If the animation is being made with insufficient data, and making incorrect assumptions, it is possible that it may mislead investigators and the public as they uncritically accept digitally altered portrayals," Eliot told CBC News.
This has been a long-standing issue with composite images or sketches, when no photo of a suspect is available, explained Tamir Israel, director of the privacy, surveillance and technology program at the Canadian Civil Liberties Association.

The Bolo Program website displays AI models of Canada's 25 most wanted fugitives, shown in this handout image, on Thursday. An AI rendering of Satinderjeet Singh, the No. 1 most wanted, is in the top left corner. (Handout/Fondation Stéphan Crétier/The Canadian Press)
"That practice has historically been a significant contributor to wrongful convictions and wrongful arrests in a way that hits more heavily on particular demographics, like Indigenous people and Black people," Israel said.
"And this AI trend of generating images raises the same sets of concerns, particularly if low-quality images were being used to generate something that looks hyper-realistic."
The AI-generated images also appear on the landing page of the Bolo website, as well as in individual entries for each person when you click through for more information.
It's only when you keep scrolling that you see actual photos of the suspects.

Satinderjeet Singh, also known as Goldy Brar, is wanted in the 2023 murder of Sukhdool Singh Gill, Winnipeg police said in a notice issued Thursday. He is shown here in undated photos provided by Winnipeg police. (Submitted by Winnipeg Police Service)
A 'proven model'
This year's most wanted list includes fugitives wanted by police services across the country, including Montreal, Toronto, Winnipeg, Edmonton and Delta, B.C.
At the top of the list is Satinderjeet Singh, also known as Goldy Brar, wanted by Winnipeg police on a Canada-wide warrant for first-degree murder in 2023. Bolo is offering $250,000 for any information that leads to his arrest. He's known to be a high-ranking member of the Lawrence Bishnoi gang.
Langlois has said that Bolo's methods have a proven track record. In 2025, for instance, 13 out of 25 of the most wanted suspects were arrested — the most of any year since the program's inception in 2018, according to Bolo's annual report.
The police units working with the program also praise it. In a news release where he commented about Singh, for instance, Winnipeg police Deputy Chief Cam Mackid said, "India is looking for him. The United States is looking for him. And now, Canada, with the proven model of the Bolo Program, is looking for him."
"The Bolo Program demonstrates the value of police services and community partners working together across Canada to extend the reach of our investigators," Niagara Regional Police Service Chief Bill Fordy said in another news release Thursday.
The Toronto Police Service noted that, since Bolo's inception, 19 of the 28 Toronto cases featured in Bolo campaigns have resulted in arrests.
But Israel, with the Canadian Civil Liberties Association, says he wonders how many false leads and accusations come from these campaigns, and what happens to the people falsely identified. This information doesn't appear to be tracked, he added.
CBC News posed the question to Bolo and has not yet received a response.
WATCH | Suspected Bishnoi gang leader tops most-wanted list:

Suspected Bishnoi gangster wanted in 2023 Winnipeg murder
November 19|
Duration
2:00
Winnipeg police say Satinderjeet Singh, a suspected Bishnoi gang lieutenant also known as Goldy Brar, is wanted in the 2023 killing of Sukhdool Singh Gill — an alleged member of a rival gang.Private entities in police work
Both experts who spoke with CBC News said there can be serious concerns when police partner with third-party companies, such as Bolo. That's especially the case when applying AI tech, says the University of Ottawa's Eliot, adding that transparency and oversight are key for maintaining public trust.
In 2021, for instance, Canada's privacy commissioner concluded that the RCMP's use of a controversial, third-party facial recognition technology was a serious violation of Canada's privacy laws. This led to major changes in policing oversight when it comes to technology, Eliot said.
RCMP's use of facial recognition tech violated privacy laws, investigation finds
Privacy commissioner to investigate Canadian Crime Stoppers breach
"The issue raised by cases such as Bolo is how these oversights apply when policing technology is being deployed by a third party working in co-operation with the police," he said.
A third-party organization may be doing so without the same set of safeguards as the police might have had themselves, added Israel. We saw this recently with the Crime Stoppers data breach, Israel noted. A CBC News investigation last month revealed that over 1.5 million Canadian tips submitted to Crime Stoppers programs over the last 25 years had been exposed.
WATCH | Crime Stoppers hack exposes personal data of Canadians:

Crime Stoppers hack exposes personal information of tens of thousands of Canadians
October 22|
Duration
3:40
A massive data breach leaked 25 years worth of unredacted personal information from Crime Stoppers, compromising the identities of thousands of Canadian tipsters. The Toronto branch of Crime Stoppers has asked the public not to send in tips until the issue is resolved."I would say it's the same thing for this kind of entity. If police were doing this there would be some rules in place for the need to be accurate, and what is being generated, and some safeguards in place for how these composite images will be used down the road," Israel said.
Bolo has good intentions, according to Eliot, who says he doesn't see any immediate red flags in the program, but he's still "quite cautious."
"We must ensure that the application of AI technology in policing is done with proper oversight, consultation, disclosure and consideration for Canadians' rights and freedoms."
KioskNews shows a cleaned-up reading view extracted from the publisher’s page — the original always lives on their site, not ours.