Report Calls for a Reset as Facial Recognition Transforms Border Crossings Around the World

Facial recognition is a highly controversial technology that is transforming border crossings around the world by fuelling an unprecedented level of surveillance & too-often racially biased automated processing of travellers. In a report released today [overview], CIPPIC documents the rapid adoption of facial recognition technologies at borders while outlining its intrusive potential and propensity for being repurposed. Facial recognition can surreptitiously identify individuals from a distance and from any live or historical image, posing a serious threat to real-world and online anonymity. Facial recognition is becoming embedded in all aspects of border crossings driven by an attempt to process travellers more efficiently and securely. But members of marginalized communities are often most heavily impacted when the technology goes awry, as some demographic groups experience far higher error rates than the general population due to lingering biases. Around the world, facial recognition systems whose creation was justified in the border control context have been repurposed for law enforcement, national security agencies, traffic safety officials, administrative agencies and even the private sector.

Our current legal framework is simply too outdated and lacking in clear safeguards to mitigate the more problematic elements of facial recognition systems. The report therefore recommends a moratorium on the adoption of facial recognition systems at our borders, and a publicly transparent reassessment of existing systems in Canada.

The report, entitled "Facial Recognition at a Crossroads: Transformation at our Borders & Beyond", was finalized with assistance from Rachel Kuchma, William Burke, Ryan Mosoff and Emily Kim. An accompanying Overview document excerpts key aspects of the core report.

Key findings of the report include:

Facial recognition technologies are inherently surreptitious and intrusive, operate with deep racial biases, and are highly susceptible to being repurposed when initially adopted in border control contexts.
Facial recognition is currently enjoying rapid adoption at border control settings primarily driven by technological developments, perceived higher levels of social acceptance in comparison to other biometrics, and the need for more efficient traveller processing.
Efficiency gains are generally achieved by automating manual travel document verification and relying on facial recognition to facilitate automation of other processes such as baggage check, customs and immigration processing and security risk assessment.
Facial recognition is rapidly becoming the biometric of choice for automating several elements of the border crossing journey, providing the essential identification component necessary for applying a range of physical and analytical automated tools to travellers. The goal is to displace other travel documents—your face will be your passport.
Efficiency gains are often overstated and fail to take into account an automated border control mechanism’s true ability to process travellers relying instead on the theoretical matching accuracy of a facial recognition algorithm while ignoring real-world accuracy challenges and related but extraneous factors.
Facial recognition is more invasive than many other biometric techniques—it retains the general biometric ability to persistently and universally identify individuals, but is able to do so far more surreptitiously and from a distance.
Facial recognition remains less accurate than other forms of biometric recognition and is persistently challenged by deep racial biases. Adoption of facial recognition systematizes these biases, with the benefits and hazards of embedding such systems at the border unevenly distributed, to the detriment of marginalized groups.
Where facial recognition is applied as a gate-keeping technology, travellers are excluded from border control mechanisms on the basis of race, gender and other demographic characteristics (e.g. country of origin). Frequently, this differential treatment will perpetuate negative stereotypes and amount to unjust discrimination.
In some border control contexts, the errors and racial biases inherent in facial recognition technologies can lead to serious repercussions, with travellers erroneously subjected to more intrusive searches, deportation, refoulement and reputation harms.
While border crossings have always been characterized by high levels of surveillance, facial recognition systems being adopted across the world are transforming ports into panopticons that increasingly extend well beyond the border by incorporating mobile devices.
Facial recognition systems adopted in border control contexts are increasingly being repurposed for a range of digital identity management, data consolidation and public safety surveillance systems. The inherently coercive nature of the border context allows for lawful and at times voluntary adoption of these systems.
The lack of clear legal safeguards allows for ongoing adoption of facial recognition technologies by border control agencies, and even by individual agents, on an ad hoc basis without dedicated lawful authorization or safeguards.
Current general legal safeguards do not provide an adequate framework for ensuring facial recognition systems are adopted in a manner that is transparent, proportionate and accountable, with sufficient consideration of the racial biases and other implications of the technology.
Canada’s adoption of facial recognition systems has been characterized by excessive secrecy surrounding the accuracy and racial bias of these systems and few clear legal safeguards to prevent systems adopted through the coercive border control context from being repurposed more broadly.