While he deserves some praise for halting the use of a controversial image search tool, Ann Cavoukian, the head of Global Privacy & Security by Design, and the former Information and Privacy Commissioner of Ontario, says Toronto police Chief Mark Saunders has got a serious problem on his hands.
“He’s got to get his house in order,” she told IT World Canada. “How can other officers and various departments use something like this without his authorization?”
Cavoukian, referring to Clearview AI, a U.S.technology company that supplies law enforcement agencies with powerful image recognition software, was surprised to find out that some Toronto Police Service members had been using Clearview AI last October.
The Globe and Mail first reported the news about TPS and its use of Clearview AI Thursday, on the heels of extensive coverage from the New York Times about the software’s use in the U.S.
“The Chief directed that its use be halted immediately upon his awareness, and the order to cease using the product was given on February 5, 2020,” police spokesperson Meaghan Gray said in an email to Global News.
A TPS department head, who did not wish to be identified, told IT World Canada shortly after the news broke Thursday that they had no idea the technology was used internally.
“I am aware of Clearview AI, but I had no idea where it was used here or for how long,” they said.
Last May, TPS confirmed reports that it was using facial recognition technology. The Toronto Star reported that Staff Insp. Stephen Harris of the Forensic Identification Services had told the publication that there were no plans to expand the TPS’s use of facial recognition “beyond comparisons to its pre-existing mugshot database.”
Clearview AI’s website says that they’ve “helped law enforcement track down hundreds of at-large criminals, including pedophiles, terrorists and sex traffickers. It is also used to help exonerate the innocent and identify the victims of crimes, including child sex abuse and financial fraud.”
Cavoukian said she has doubts about those claims.
“They don’t mention the false positives,” she said.
Clare Garvie, a researcher at Georgetown University’s Center on Privacy and Technology, told the New York Times that no data exists that proves Clearview AI is accurate.
“The larger the database, the larger the risk of misidentification because of the doppelgänger effect. They’re talking about a massive database of random people they’ve found on the internet,” she was quoted saying.
On its website, it says, “Clearview’s image search technology has been independently tested for accuracy and evaluated for legal compliance by nationally recognized authorities. It has achieved the highest standards of performance on every level.”
No additional detail is provided.
One of the pull quotes on Clearview’s website says it’s from a “Detective Constable in the Sex Crimes Unit” from a Canadian law enforcement agency.
“Clearview is hands-down the best thing that has happened to victim identification in the last 10 years. Within a week and a half of using Clearview, [we] made eight identifications of either victims or offenders through the use of this new tool,” the pull quote reads.
Requests for access can be filed through Clearview AI’s website in a matter of minutes.
Cavoukian said it’s sometimes normal for large organizations to lose track of innovations that are rapidly being tested, but it’s scarce to read about cases where the top of the food chain is entirely unaware of what’s happening below.
“Clearview AI, they scrape billions of images from social media and other platforms without any permission and no notice … it’s appalling.”
At the very top of its website, Clearview AI says it only scrapes “public information only” and that its tool is for “search, not surveillance”.
However, Facebook and other social media sites prohibit people from scraping users’ images, meaning Clearview is violating the sites’ terms of service. Even the company’s founder, Mr. Ton-That, an Australian techie and onetime model, according to The Times, admits this.
“A lot of people are doing it,” Mr. Ton-That was quoted. “Facebook knows.”
