AI-generated material found in one

AI-generated material found in one-third of P.E.I. RCMP online child abuse cases

RCMP report a 'notable increase' in AI-generated child sexual abuse material on the Island

A third of all Prince Edward Island RCMP investigations into online child sexual abuse now involve material generated by artificial intelligence, RCMP say.

In an email, P.E.I. RCMP senior communications adviser Scott Ferris reported a "notable increase" in AI-generated child sexual exploitation material in areas under RCMP responsibility on the Island.

Ferris called it a "safe estimate" that 33 per cent of RCMP online child abuse cases now involve AI-generated material. Estimates exclude Charlottetown, Summerside and Kensington, which have their own police forces.

"[AI] is being utilized now to generate images or to manipulate images of young people," said P.E.I. RCMP Const. Tim Keizer. "It’s becoming increasingly difficult to distinguish because AI-generated images are very realistic."

Canadian law treats AI-generated child sexual abuse imagery the same as real depictions. Section 163.1 of the Criminal Code prohibits any material depicting a minor for a sexual purpose.

"There’s no difference," Keizer told CBC News. "There is a charge just the same as if it wasn't done via AI."

In a July email, Cpl. Gavin Moore, P.E.I. RCMP media relations officer, said roughly two RCMP cases each month involving child sexual abuse imagery result in charges. Due to factors like insufficient evidence, he said this figure likely represents less than half the number of cases received.

Police say 19 victims identified in AI child sexual abuse image investigation

'Dad, I screwed up': A year after their son's death, P.E.I. family speaks out about sextortion

RCMP estimate that AI models are used in around a third of cases to either generate images and videos of non-existent children for sexual purposes or to create "deepfakes" — digitally altered versions of real people.

Deepfakes are often used in sextortion cases, where criminals threaten to leak sexual images to extort money or explicit content from victims.

Moore told CBC News that P.E.I. RCMP had 26 reported instances of sextortion last year. So far this year, Summerside police identified one sextortion case involving AI-generated child abuse material, along with another instance of suspected AI-generated child sexual exploitation imagery.

A recipe for nabbing child predators: Take tips from online platforms, add 'dedicated investigators'

Charlottetown police identified one instance of AI-generated child abuse material so far this year.

Keizer said young people and parents should be educated on the potential dangers of sharing images online.

"We have to be aware that these images can be taken from the internet and utilized and manipulated in a wide variety of ways," he said. "We have to be socially responsible with our technologies."

Even when AI-generated material doesn't depict real children, the harm remains significant, according to Jacques Marcoux, director of research and analytics with the Canadian Centre for Child Protection.

"A lot of these AI-generated images are also based off of the likeness of real kids in our communities," he told CBC News.

"[Some AI models are] a vast scraping operation that indiscriminately vacuums up images all across the internet, and there's a lot of child sexual abuse material on the internet that hides in plain sight on pretty mainstream platforms that a lot of Canadians use."

Although the growing realism of AI content doesn’t change anything from a legal standpoint, given the Criminal Code definition, Marcoux still stressed the importance of specifically tracking AI-generated exploitation.

"It’s important to keep tabs on it as much as possible because that gives us a sense of the extent to which this problem is growing," Marcoux said.

"There’s no doubt that AI is also growing an interest in demand."

Joseph Watt is a reporter and associate producer with CBC P.E.I. He holds a master's degree in International Journalism from City University of London. He can be reached at joseph.watt@cbc.ca