City police avoid open-source AI technology
Facial recognition software that uses social media images raises ethical issues
A Calgary police officer says using open-sourced data such as social media with facial recognition technology is “fraught with danger.”
The comments were made in reference to controversy surrounding the use by Canadian police forces of an American company’s facial recognition software that created a massive database of open-sourced images taken from platforms such as Facebook.
Staff Sgt. Gordon Macdonald, of the criminal identification section of the Calgary Police Service, said using such a database isn’t an option for the local service.
“At the moment, the Calgary Police Service has no intention of going near any facial recognition software that would utilize images we have no control over or we can’t verify,” said Macdonald.
“As clever as it sounds and advantageous, you always have to have the ethical perspective on things. When you don’t have ownership of these images, you don’t know who has posted them, how they’ve posted them.”
Calgary police currently use facial recognition software to compare photos and videos, such as CCTV images of persons of interest, with their mug shot database of more than 350,000 images taken under the Identification of Criminals Act. Potential matches are then reviewed by a trained facial recognition technician.
Questions regarding the use of facial recognition technology by Canadian police forces surfaced after a New York Times investigation into a company named Clearview AI.
The story detailed how the company created a large database of open-sourced images taken from platforms such as Facebook, which could be used by police to identify perpetrators and victims of crime. The New York Times investigation also reported that Clearview AI said Canadian forces were using the technology.
The story prompted questions by the Ottawa Citizen, which reported that RCMP would not answer if they used the Clearview AI technology. Similarly, the Ottawa Police Service would not specify the facial recognition software they used during a pilot that ended in March 2019.
Canada’s federal privacy watchdog has since launched an investigation into the Canadian use of facial recognition technology supplied by Clearview AI. It will be a joint investigation with provincial counterparts in Alberta, British Columbia and Quebec to determine whether the organization’s practices comply with Canadian privacy laws.
“As an organization, we wouldn’t be interested in it no matter the benefits it purports to bring. It’s just so fundamentally and ethically unsafe to start using that as a means to obtain some form of identification,” said Macdonald.
“It’s far better to go through our own photographs that we’ve obtained and can verify who these people are.”
The Edmonton police force said this month it will implement similar facial recognition technology as Calgary’s to match suspects with known offenders. Edmonton’s pro
When you don’t have ownership of these images, you don’t know who has posted them, how they’ve posted them.
gram is expected to roll out later this year.
Macdonald hopes the two police forces can eventually link their programs to cross-reference known offenders from both major Alberta cities.
Calgary police were the first police agency in Canada to utilize facial recognition technology in 2014.
Macdonald said a year of research, review and evaluation of facial recognition tools resulted in a contract with the NEC Corporation of America to use the Neoface Reveal software.
“There’s a demand for public safety that pushes the use of more
AI stuff, but I think back then, in 2014, there was not enough public knowledge about facial recognition,” said Macdonald. “It was still a novelty.”
He said there are still some misconceptions from the public about the use of facial recognition software, but if it is used responsibly and ethically it has merit to save time and resources.
Calgary police have since upgraded to the Neoface Reveal 2.0.
It is the most advanced version of NEC Corporation of America’s software in Canada and Macdonald said the upgrade has proved effective when compared to the previous model. For example, eight CCTV images the original system couldn’t search due to low quality yielded high-value matches in the upgraded system.
Macdonald said using facial recognition software has proven successful for the force, and helped solve some high-profile cases.
“Everyone is looking to solve crime quicker and faster, but you have to be cautious about how you go about that,” he said. “A system like this is really effective. I think this program is a great way for us to show how we are innovative and efficient at the same time, but also cautious as well to how we apply it.”