India, in recent years, has been experimenting with Facial Recognition Technology (FRT). The technology is claimed to have been used for tackling crime, and providing services- digi yatra, life pensioners scheme etc. FRT has also been deployed, quite problematically, to identify protestors in some cases. This rampant push for FRT has happened despite there being no legal framework to govern FRT and check its misuse. And now, it seems to have made its foray into elections. As per a recent report published in MoneyControl.com Election Commission of India (ECI) was all set to pilot FRT in Karnataka Assembly Elections.
As per an Election Commission official quoted in the story, a voter would need to download the Chunavana app, provide EPIC and mobile number to get an OTP for verification and then upload a selfie. On reaching the polling booth, a voter would undergo facial recognition scanning. If the selfie matched the photo in the database of ECI, the voter could proceed with voting, no other document required. However, after the publication of the report, CEO Karnata clarified that the proposal was still under deliberation and the same would likely not be implemented. Ultimately the pilot didn’t happen for the Karnataka elections which were held on May 10th 2023. The problem with the app, as pointed out by a Twitter user, was that the app could share the data of the voters to third parties making voters vulnerable to exploitation.
This was, however, not the first time that FRT was proposed in elections. In 2020, the Telangana State Election Commission issued a circular informing that FRT would be used in Municipal elections in 10 polling stations of Komapally Municipality of Medchal Malkajgiri District. A report noted that the app yielded an 80% accuracy rate and in some cases, it could not work due to poor lighting or old pictures of voters on their EPIC.
FRT in elections: from Afghanistan to Komapally, Telangana
The first known use of FRT in elections was in Afghanistan’s presidential elections in September 2019. It was claimed that this would prevent fraud and corruption in the elections as in many places women did not vote and the male members of the family instead cast votes in their places. With the announcement of FRT being used in the elections of 2019, many conservative families became suspicious towards the process and refused to participate in the elections and women termed this as an infringement of their privacy.
The process itself was far from perfect as there were long queues waiting outside the polling booths and it proved to be tedious as in some cases, the process was taking ten minutes to identify the voter.2 Another issue was that it prevented the already underrepresented Afghan women from participating in the election process and their sensitivities were not taken into consideration.3
India piloted FRT in the election process for the first time in January 2020 in Telangana. This was the second known project in the world, after Afghanistan.4 A research paper that conducted an empirical study of the political consequences of FRT in the municipality election noted that the voter turnout was lower in the polling stations with FRT. Particularly, the Muslim voters stayed away due to heightened concerns about government surveillance using facial recognition, especially in the light of protests against CAA and NRC. The largest Muslim political party in South India, AIMIM did not field its candidates from the locality and one of its MLCs wrote a letter outlining issues with FRT, such as privacy, legal consent and misuse of the photo and creation of a large database of facial contours and features.
FRT vis a vis Right to Privacy and Puttaswamy Judgment
The Supreme Court of India in Justice KS Puttaswamy (Rtrd) vs. Union of India (2017) 10 SCC 1 recognised privacy as one of the fundamental rights and laid down a test to measure whether a certain executive action is privacy respecting or violating.
The first prong of this test is legality - any executive action which can potentially violate privacy must be backed by a law. There is no explicit law that bestows the government or investigative agencies with the power to use FRT for any purpose. In all the use cases of FRT mentioned in this article, there were executive orders for their conceptualisation and implementation. However, as per the Aadhaar Judgment of the Supreme Court, an executive order doesn’t meet the test of legality as it requires a law made by the legislature which must also be just, fair and reasonable. Puttaswamy further states that there must be a legitimate state aim at the basis of a law which imposes restrictions on or violates the right to privacy. It must guarantee against arbitrary state actions. The restriction or violation proposed must also be proportional to the objective sought to be achieved.
The use of FRT in elections fails the other prong of Puttaswamy test, of necessity- to check if there are no other less restrictive means to achieve the same ends. Is it to prevent electoral frauds? If so, how rampant are electoral frauds and how have conventional methods failed? Has any assessment done of the risks associated with the process? How will the sensitive data of voters gathered through FRT be handled? And equally importantly, how will the errors in the database be addressed which might actually lead to disenfranchisement of genuine voters?
As reported in a research paper the deployment of FRT over large segments of the population for varied reasons may be regarded as disproportionate to the objective sought to be achieved.7 Furthermore, in the absence of a data protection regime, the issues such as privacy of the individuals, purpose limitations, accountability of the state, and the grievance redressal mechanism in cases of the breach of data remain unaddressed, therefore jeopardizing the safety of the individuals.
Conclusion
There is no clarity under what law the Election Commission of India is exercising this power to deploy FRT in elections and what safeguards will be put in place by ECI to protect the data of the citizens from misuse.
As pointed out by several digital and human rights activists, the process of FRT in elections would lead to mass disenfranchisement as many genuine voters would be left out of the electoral rolls due to the mismatch of their photos with EPIC. Another concern is surveillance and profiling of the citizens and this fear may deter voters from exercising their right of universal adult franchise.
While there are 109 countries currently using FRT for different purposes, there are countries which have banned its use in their respective jurisdictions as well.8 France and Sweden have banned FRT in schools. Earlier this month on 11 May, 2023, two committees of MEPs overwhelmingly endorsed sweeping new rules on artificial intelligence. Now the entire EU will vote in June and once approved the regulation may become the world’s first comprehensive AI law banning predictive policing and facial recognition9. The cities in USA such as San Fransisco, Somerville, Massachusetts, Oakland, and California have banned the city departments including the police from using FRT.10
In the absence of a data protection law, India should not proceed with the use of FRT for any purpose, particularly in elections. As the situation stands today, the risk of FRT outweighs the possible benefits. If anything, the government must initiate a transparent dialogue to objectively understand and assess the pros and cons of FRT before embarking on its use in any domain.
Endnotes
1. A voter registration pilot exercise using facial recognition technology and iris scanning was conducted between 2006 and 2008. Margie Cook, Afghanistan Electoral Support: June 2008–January 2010, March 2010, p. 20. However, it was unsuccessful. For example, the facial recognition pilot ran into problems caused by the prevalence of “burkas and beards” in Afghanistan, resulting in high failure rates. Senior elections advisor, SIGAR interview, September 26, 2018. Indeed, one of the challenges of using biometric data to register, identify, or verify voters in Afghanistan is the cultural sensitivity to women showing their faces in public. The IEC feasibility study noted that, while it degrades the quality of the biometric database, consideration should be given to allowing women the option not to have their photograph taken, in order to avoid mass disenfranchisement.
2. https://www.reuters.com/article/us-afghanistan-election-technology/biometric-machines-in-afghan-vote-improve-after-last-years-glitches-idUSKBN1WD0DM
4. Facial Recognition Technology and Voter Turnout, Feyaad Allie https://static1.squarespace.com/static/62a0e15f0bc6e3095c1ff089/t/63111cd2b13581559ef03680/1662065879995/FRT_JOP_final_FeyaadAllie.pdf
5. KS Puttaswamy vs. Union of India (2017) 10 SCC 1, page 505
6. ibid
7. Facial Recognition Technology in India, Centre for Internet Society, India