After first denying it, Toronto police confirmed this week that some officers have been using a powerful new facial recognition technology known as Clearview AI.
According to published reports, Toronto police chief Mark Saunders only became aware officers were using the technology less than two weeks ago after a reporter from the Star asked about it. He has ordered them to stop using it until the police department can conduct what a police spokesperson refers to as a “fullsome review.”
The facial recognition technology is capable of scraping the internet and turning up detailed information about a person based on a photo.
Welcome to Toronto’s surveillance state.
Clearview AI is not the only technology used by the Toronto police to surreptitiously record and identify potential suspects and intercept cell phone communications without requiring a warrant.
Toronto police also use a device known as StingRay, which allows police to track people’s whereabouts and communications through their phones.
Soon after the former Ontario government of Kathleen Wynne decided to regulate the practice of random street checks, popularly known as carding, Saunders declared that he would have to “find an Option B.’” Surveillance technology seems to be that option and it seems to have been deployed quietly.
The Toronto Police Services Board (TPSB) does not appear to have been kept in the loop on the use of the StingRay technology. Torontonians only became aware of this snooping equipment through media reports. Even after media reports, members of the board did not ask any questions – at least not at its regularly scheduled monthly public meetings.
On the other hand, the board did receive several rather sketchy but reassuring reports between 2017 and 2019 on the purchase of facial recognition technology. This was preceded by a pilot project reportedly conducted between September 2014 and September 2015. But virtually no information or any independent evaluation has been provided by the police department.
There were no public consultations or efforts to develop a policy prior to approving the purchase of equipment from the U.S.-based NEC Corporation. The same province that had insisted on bringing in a mandatory regulation for carding, gave the Toronto police $1.6 million from its policing modernization fund to buy devices.
There is also no publicly available information from the force about how devices are being used or how the collected data is stored and accessed. There is no information either about what is being done with data captured on innocent bystanders – despite significant research to show that the technology is not wholly reliable. It cannot always recognize facial images captured from different angles or in different lighting conditions. There is, therefore, a high risk of false matches, especially of Black and other people of colour. Questions about its reliability are what caused the city of San Francisco to ban the use of facial recognition tech by police and other law enforcement agencies.
In Toronto, the board’s lackadaisical handling of the police service’s interest in the technology stands in stark contrast to its response to the introduction of CCTVs almost a decade ago.
At that time, a series of meetings were held on CCTVs. Members of the public were given the opportunity to ask questions. Based on these sessions, the board developed a policy governing even the pilot project and required an independent evaluation before any further work could be done.
Now, the service is comparing people’s faces with images in its mugshot database without board oversight or public scrutiny. There is only the assurance in chief Saunders’ report on facial recognition technology to the board’s May 30, 2019 meeting. It states that the tools being used do not scan crowds and are “only used lawfully, in accordance with the Criminal Code.”
Saunders also assures in his report that the police service has addressed the privacy concerns in consultation with senior officials in the Crown Prosecutor’s office, the province’s Ministry of the Attorney General and the force’s in-house legal counsel.
Normally, it is Ontario’s Freedom of Information and Privacy Commissioner who provides advice and helps develop policies and procedures to protect people’s privacy from police gathering of information. Saunders’ report does not explain the departure from this practice.
Nor does it offer any concrete evidence to support the claim that the system has proved “very useful, effective and efficient in identifying potential suspects.”
According to the report, Toronto police conducted 1516 searches between March 22, 2018, and December 31, 2018, using some 5,000 still and video images of “suspects” obtained through various sources. These images were compared with the 1.5 million mugshots stored in the police database. The “potential” suspects were then manually analyzed by police personnel.
However, Saunders cannot tell precisely how many arrests resulted from the use of this technology operated by a dedicated team who, prior to the pilot project, received “Face Comparison and Identification training” at the F.B.I.’s Criminal Justice Information Services Division in Clarksburg, West Virginia.
There are several reasons why this should be of serious concern to the public and why the board must revisit its extremely perfunctory decision-making on this issue.
Despite research showing that the software is not totally reliable, there are serious privacy issues.
For example, it is not clear how still and video images obtained through different sources are selected. The May 2019 report does not spell out the process. Are these images chosen randomly based on a superficial resemblance? Furthermore, what is done with those images that are rejected in the process? There is a similar problem with the random and broad nature of the secret recording of cell phone communications.
Also, the use of the technologies raises questions about the sharing of private information among police forces as criminal activities that cross local jurisdictions are investigated by joint task forces and officers from different police departments.
At around the same time as Toronto began its pilot project, the RCMP was looking into adding facial recognition technology to its arsenal of identification systems which, until then, was chiefly based on fingerprints.
The RCMP has been working to replace its existing nationwide Automated Fingerprint Identification System, or AFIS, with a new system that, according to Motherboard, “could analyze and capture faces, fingerprints, palm prints, tattoos, scars, and irises.”
With federal legislation making it easier for federal agencies like the RCMP and Canada Border Services Agency to share information, the interest is in a system that enhances what in Canadian and U.S. law enforcement circles is called “interoperability.” The RCMP wants a system that improves interoperability not only among Canadian law enforcement but also with its international partners, such as the FBI.
As Canada’s largest municipal law enforcement service, how will the Toronto police remain immune from these larger developments? Just as the current AFIS database is national, will its replacement – a far more ambitious biometric system – be national too? What will be the scope of the use of this technology?
Among Saunders’ early justification of carding was its usefulness in dealing with national security. That is a very different area of policing than a criminal investigation, which according to Saunders, is the only purpose for which facial recognition technology is being used in Toronto.
What is the guarantee that the technology will not be used to deal with border security issues or expressions of civil dissent, such as protests and demonstrations?
The Toronto Police Services Board has an obligation to exercise the oversight that it has failed to do.
Alok Mukherjee served as chair of the Toronto Police Services Board from 2005 to 2015.
@nowtoronto
