If robust policies are to be developed for the use of facial recognition technology across Latin America, concerns around human rights must be addressed and safeguards built against potential abuses.
While there is widespread agreement on the shortfalls and perils that the use of facial recognition poses, there are two distinctive positions on how the technology should be handled.
First, there are those who side with authorizing the use of the technology in public places for law enforcement purposes, provided that strong safeguards are in place. At state level in the US, and in the debate around the forthcoming AI Act in the EU, policymakers have worked to legislate based on the notion that facial recognition is a valuable tool when used to protect public security, and, as such, have sought to regulate the technology by allowing its exceptional use in public places, requiring judicial or third-party authorization except in cases of specific emergencies. For Latin America, where the technology is already deployed, such an approach could be set in motion through the establishment of moratoriums that allowed the use of the technology to be paused until proper safeguards are put in place.
On the other side of the debate, privacy and data protection watchdogs and civil society organizations maintain that facial recognition is inherently at odds with fundamental rights such as privacy, and that its use for law enforcement in publicly accessible spaces is unacceptable in democratic societies. This view has been epitomized by the city-level bans rolled out across the US. It echoes the stance taken in Latin America by human rights groups and digital rights activists, who are calling for outright bans throughout the region.
In Latin America, where the technology is largely deployed without proper safeguards, both policymakers and human rights advocates must ask themselves what the most realistic routes might be to move beyond the current worst-case scenario. As regulation in Western jurisdictions appears to be headed towards exceptional, authorized use with strict safeguards on that use, placing a moratorium on the technology while those precautions are put in place seems a feasible alternative for Latin America. However, this is not a bullet-proof approach, and may not be easily applicable to non-consolidated democracies and nations with governments tending to authoritarianism.
Indeed, proponents of facial recognition bans in Latin America have strong and valid arguments about why the region may be unprepared for ‘middle-of-the-road’ solutions. First, they highlight that state institutions have often performed poorly when enforcing safeguards around surveillance practices. Brazil is one such example: its wiretapping law (Law No. 9,296, enacted in 1996) has been lauded for meeting international standards and safeguards. Yet, far from being an instrument for exceptional use, security forces are regularly and easily granted wiretap permits, as evidenced by the authorization of an extraordinary 21,925 phone tapping cases during protests against Brazil’s staging of the football World Cup in 2014. Enforcement capacity may also be challenging for smaller Latin American states, where institutions are relatively fragile and may not have the resources or political leverage to properly monitor the adequate deployment of facial recognition technologies.
Another point of concern among supporters of a comprehensive ban is the potential amplification effect, inherent to facial recognition technologies, of algorithmic and data bias, which is likely to exacerbate existing inequalities and reinforce structural discrimination. Pro-ban activists suggest that marked racial, gender and income inequalities across the Latin American region would render the deployment of facial recognition particularly risky for vulnerable populations. For example, some of the limited data available in Brazil indicates that people of colour are disproportionately targeted, as compared to white Brazilians. While available data on the Buenos Aires deployment is not disaggregated by race, gender or income level, the use of facial recognition to tackle petty crime clearly sets out to criminalize vulnerable populations in Argentina, where poverty rates stood at 40 per cent in 2021.
Pro-ban groups also point to the history of military dictatorships in Argentina, Brazil and other Latin American countries. These regimes share a regrettable track record of incarcerating political dissidents and engaging in political assassinations and kidnappings –a sobering reminder that civil society must stand for the protection of open public spaces that are free of surveillance. The human rights abuses experienced under military rule are a fresh example of how the potential misuse of facial recognition for state surveillance can incur significant human and political costs. While Latin America has not yet seen major attempts to abuse the technology, China’s alleged use of facial recognition to track Uighur populations in the Xinjiang region is a worrying example of such misuse.
The value of moratoriums has also been questioned by civil society actors. Representatives from European Digital Rights (EDRi), an association of European civil and human rights organizations, claim that waiting for facial recognition systems to improve performance and tackle bias will still not address potential infringements of privacy. The technology would continue to erode anonymity, transform the public space and undermine the very foundations of open and free societies. Offering strong safeguards to limit the use of facial recognition in publicly accessible spaces would only normalize some degree of mass surveillance. This could be manageable in Western jurisdictions with strong institutions and robust enforcement capacities: however, it may not be an acceptable standard across either the region or the Global South, where normative safeguards may be weak from the outset.
As multiple countries are exploring how to approach the use of biometric and AI technologies, this is the right time – in Argentina, Brazil and Latin America more broadly – for an open debate about how to move forward with regulating facial recognition. A central challenge will revolve around whether to attempt to put ‘the genie back in the bottle’ through the imposition of outright bans (as framed by Carly Kind, Director of the Ada Lovelace Institute) or to focus on building the proper safeguards to prevent abuses deriving from ongoing uses of this technology. But could such protections prove sufficiently robust in Latin America?
While arguments in favour of a ban are compelling, it is also clear that no safeguards would be a worse outcome than some safeguards. Constructing adequate protections – such as having strict internal and external oversight mechanisms over the institutions in charge of conducting facial recognition; offering proper avenues for redress in cases of discriminatory uses of the technology, or false positives; and opening data on the performance of facial recognition systems to assess effectiveness and bias – may test local institutions in Latin America, but these would constitute important steps in moving regional regulation and facial recognition implementations in the right direction.
In the case of Buenos Aires, where the use of live facial recognition was fully operational until April 2022, policymakers need to urgently reassess the potential human rights impacts of such practices and acknowledge the extent to which current safeguards offered in Argentina are far behind acceptable, minimum standards in the EU, UK and US. In the case of Brazil, where the public debate around the use of biometric technologies has been more sophisticated, the need for regulation to guide facial recognition deployments and pilot projects is becoming ever more pressing. Beyond law enforcement-driven uses of the technology, regulation must extend to other problematic applications of facial recognition within Brazil, such as in the distribution of social benefits.
Recommendations
Some additional recommendations can be drawn for other Latin American policymakers exploring strategies to regulate facial recognition technologies:
Facial recognition regulation should be anchored on human rights. Latin American countries share a strong tradition of engagement with regional and international human rights processes, and human rights standards are widely accepted in the continent. As the region continues to explore how human rights standards apply to emerging technologies and digital environments, the regulation of facial recognition technologies must be anchored in these guiding principles and must uphold regional commitments to the protection of individuals’ rights to privacy, to freedom of opinion and expression, and to freedom of peaceful assembly and association, as well as rights to non-discrimination and due process. In particular, any regulation designed to enable exceptional law enforcement-driven uses of facial recognition in public spaces must build robust safeguards to protect said human rights. There are some internationally agreed standards on what constitute proper safeguards. These include: (a) ensuring transparency around the deployment and use of facial recognition technologies, with the regular reporting of performance data – as validated by external oversight – and putting mechanisms in place for a swift response to freedom-of-information access requests; (b) rendering procurement procedures for the acquisition of surveillance technology more transparent, with proper disclosure of information on technology providers and the features of the technologies employed; (c) guaranteeing data security and adequate data handling procedures, ideally as guided by national data protection legislation, if in place, or following international standards; and (d) guaranteeing internal and external oversight of specific technology deployments, including mechanisms for redress for those affected by wrongful identifications. In addition, private companies engaged to provide the technology or run the software on behalf of governments should abide by human rights guidelines as specified in the UN Guiding Principles on Business and Human Rights. The establishment of specific safeguards and mechanisms for oversight that adjust to local realities may be discussed and agreed by joint working groups coordinated by implementing authorities, with the participation of policymakers, security forces, technical specialists, civil society, privacy watchdogs and ombudsman agencies.
Authorized uses of facial recognition must steer clear of ‘no-go’ zones. International debates on the use of facial recognition and emerging regulation in Western jurisdictions are beginning to generate some valuable guidelines to deal with the technology. Strong agreement has emerged around what can be considered as no-go zones for law enforcement uses of the technology, which are those where potential drawbacks to human rights exceed potential benefits. This translates into enabling the exceptional use of facial recognition technologies in public spaces only in the case of serious criminal offences or life-threatening emergencies, and excluding what is currently authorized in the case of Buenos Aires: the use of the technology for persecution of petty crime. The regular use of live recognition – which until April 2022 was a regular practice in Buenos Aires – is also emerging as a no-go zone, with regulatory frameworks only allowing its deployment under exceptional circumstances. Requiring judicial or third-party authorization from applicable government agencies is also becoming established as a widely accepted standard, serving as a crucial safeguard to prevent abuses in the use of the technology. This has been seen, for example, in legislation in parts of the US and in the EU’s proposed AI Act. Different standards emerge when it comes to judicial or external authorizations. In the US, the debate has split between those who favour the application of probable cause as the required standard and those who favour more broad uses of facial recognition when relevant for the investigation. While more stringent standards make for more robust safeguards, what is clear is that the indiscriminate use of facial recognition in public spaces and its authorized use to identify minor offenders are both unacceptable, disproportionate applications of the technology – with the negative human rights impact of applying the technology outstripping any potential benefits or gains.
Facial recognition regulation should be sensitive to the local context. While regulatory frameworks from other jurisdictions may provide guidelines and innovative methodologies—such as Europe’s AI risk-assessment approach – the adoption of ‘copycat’ legislation should be avoided, and the design of domestic laws should take into consideration local contexts and limitations. Particular attention is required around enforcement capacities to devise feasible regulatory frameworks that are applicable to the local context and the consideration of the potential impacts of algorithm bias in reinforcing structural discrimination along racial or class lines in Latin America.
States need to engage in thorough public conversations about the use of facial recognition. Policymakers and regulators, particularly those dealing more closely with the application of technologies for public safety, must become more conversant with the potential human rights impacts of emerging technologies. They should also strive to generate open opportunities for debate prior to any type of technology deployment. The two public audiences organized in 2019 by Brazil’s National Congress and the federal public ministry, and the current consultations in the country’s federal senate, on the drafting of an AI regulatory framework, are examples of how such debate may be accomplished. The EU’s multiple public and expert consultations on the AI Act also provide a model for how this level of public dialogue may be enabled with relevant stakeholders. In addition to formal consultation processes, debates should also extend to the general public, and in particular, to populations that may be affected by specific technologies. Civil society and human rights groups have contributed to such efforts and need to be actively engaged. This may be accomplished, for example, through town-hall meetings, which are uncommon but not unprecedented in modern Latin America.
Latin American governments still have a long road ahead of them if they are to move towards the establishment of adequate regulation to deal with the use of facial recognition technologies. This will require regional policymakers to actively engage with security forces in devising adequate normative guidelines and to explore the adoption of technologies that effectively serve the public interest, while protecting fundamental rights that are essential to democratic societies. More in-depth policy analysis will be required if the use of biometric and AI-based technologies is to be limited, and national strategies for AI will need to be developed. Most importantly, Latin American governments which have allowed the widespread adoption of facial recognition technology must acknowledge the potential risks associated with its use and enact proper regulation to prevent such abuses.
