By Ameen Jauhar
The Monsoon Session of Parliament saw the withdrawal of the Personal Data Protection Bill, 2019. Against this background, recent media reportage of the Delhi police increasing its scope of reliance on facial recognition technologies (FRT), is disconcerting at best, and reckless and arbitrary at worst. Fundamentally, the continued deployment of FRT on the pretext of efficient policing (commonly adopted as raison d’être by law enforcement in India and abroad), exposes a lack of clear understanding of the technology at play, and its many limitations. FRT has witnessed a global pushback, especially against its use by law enforcement agencies. Issues range from design flaws (low accuracy, especially among women and darker-complexioned individuals), to legal and constitutional risks of mass surveillance and impingement of individual freedom and privacy.
Yet, there have been reports of the Delhi Police utilising FRT to arrest and charge individuals, most notably in the 2020 Delhi riots’ case. To worsen matters, in a separate RTI inquiry, the Delhi police responded that while deploying FRT, it considered an 80% accuracy as a positive match. Cumulatively, these two points present a brewing recipe for disaster. Let’s take each of these points separately.
One of the principal arguments against FRT is with respect to its accuracy flaws. Even proponents of the technology cannot disprove the significant error rates that even the most sophisticated FRT algorithms have suffered from. For instance, a couple of years back, infamously, an FRT algorithm misidentified Michelle Obama as a man, and predicted inaccurately when it came to sitting members of the US Congress. Similar instances since then have erupted globally, and in some cases, inaccuracies have been as high as for 75% of the matches.
Also read: Linking rural producers with value chains
While inaccurate outcomes certainly undermine the efficacy of this technology, the more detrimental impact is on civil rights of individuals, particularly when the usage is by law enforcement. Globally, there have been reports of states relying on FRT to suppress legitimate protests (as witnessed in 2019 during the Hong Kong pro-democracy protests), and in some cases, target specific individuals. The latter was quite evident in the arrest of a black man in New York in 2020, during the peak of the Black Lives Matter movement, which was based on an inaccurate facial match.
These risks, while reported largely from other jurisdictions, are also erupting at home. At the Vidhi Centre for Legal Policy, we have been working on the risks of FRT systems in India, specifically within policing and law enforcement. Our papers discussed the risks of such technologies given the lack of adequate legal checks and balances, and an utter lack of transparency around the use of such systems by local police agencies. One of our papers specifically made a case study of the ongoing efforts of the Delhi Police to integrate FRT within its policing. Based on our empirical research, we determined that due to inherent biases in local policing, the reliance on this technology is likely to arbitrarily target minorities and other vulnerable groups based in certain neighbourhoods of Delhi.
In a strict legal assessment, the use of mass surveillance technologies like FRT are arguably in violation of the Supreme Court’s ruling in Puttaswamy (the privacy judgement), especially in the current statutory vacuum. But beyond privacy, there are procedural concerns as well. First, reports are suggesting that FRT inputs will be presented as evidence to charge and potentially sentence an accused. However, the current provisions of the Indian Evidence Act, while dealing with electronic evidence, do not recognise algorithms, or algorithmic output, which is what current FRT systems are. Hence, the exact validity and even the legality of such evidence is impugnable. Then there are basic due process requirements which are well established in Indian constitutional jurisprudence.
The fact that currently no legislation (either the Criminal Procedure Code or other legislation governing policing), sets out the process of using FRT, its deployment is left to the arbitrary discretion of the executive or law enforcement. In fact, this is the very grounds of the first constitutional challenge to the use of FRT, currently sub judice before the Telangana High Court.
Lastly, there is concern around what remedies exist for individuals subject to criminal prosecution based on this evidence, as is now happening in Delhi. While India does not have a formal tort law statute, police excesses, especially when violative of fundamental rights, have long been dealt with stringently. The Supreme Court has awarded compensatory, and even punitive damages where police action has affected the right to life under Article 21 (like Rudul Shah, Bhim Singh, and Nilabati Behera). However, in the absence of a formal grievance redress mechanism, these disputes become entangled in prolonged litigation in our infamously tedious and slow justice system and may take years in actual remedial measures being afforded.
What results through this entire process is the likely arbitrary, and arguably unconstitutional use of a mass surveillance system by the police, lacking transparency, legitimacy in law, and without fair checks and balances in place. A carte blanche.
The writer is Lead, Centre for Applied Law & Tech Research (ALTR), Vidhi Centre for Legal Policy.
Views are personal