Drones: Robot eyes on racialized migrant bodies

A joint Border Forensics and Human Rights Watch investigation revealed that the Libyan Coast Guard (LCG) pull-backs are facilitated by plane and drone surveillance provided by the Frontex (the European Border and Coast Guard Agency) to the LCG about migrant boats (Sunderland & Pezzani, 2022). Migrants who are returned to Libya are enslaved, extorted, tortured, raped and killed (UNHRC, 2023); among them, Daoud (pseudonym), a student activist from Darfur, Sudan explained, ‘When I was in Libya, I was like in 17th century, not the present’.1 Drones are increasingly used in migration control operations and gaining new functions with the integration of Artificial Intelligence (AI). Nevertheless, there has been limited critical social scientific engagement around the connections between drones and migration. This commentary situates drones within the trend of militarization of migration controls, which carries significant surveillance, ethical and human rights implications, particularly for racialized migrants.2 Drones have emerged within, and significantly accelerated, the process of militarization of migration controls, which entails not only the transfer of military technology but also military rationalities to target migrants (e.g. Csernatoni, 2018). Pioneered by the US military in the post-9/11 context, drones have been heavily deployed in warzones by militaries to surveil, kill and destroy enemy targets. They have later been transferred to civilian applications, including urban policing and migration control. Even though planes have long been used in migration control,3 drones are increasingly preferred for their cost-effectiveness, speed, ability to operate in difficult weather, vast reach into difficult terrain and the low risk they pose for human operators (e.g. Lutterbeck, 2022). Drones have high levels of interoperability, can connect with other surveillance systems (such as satellites, databases and ground surveillance systems) and provide real-time video streaming of migration movements by various optical and thermal sensors, and radars, thus providing ‘situational awareness’. Following a militarized logic, drone surveillance typically results in targeting migrants as ‘threats’ along the borderzones, informing ground actors and pre-empting their entry, thus contributing to the neutralization of the ‘threat’ of migrants/migration. The data is also used for ‘risk’ assessment to understand migrant journeys and prepare future pre-emptive strategies. The ‘verticality’ (Weizman, 2007) of drone surveillance resembles Foucault's well-known metaphor of the Panopticon tower as the model of modern surveillance, which disciplines subjects under its hierarchical gaze (Foucault, 1991). But unlike the Panopticon, drones are mobile and dispersed (or ‘liquid’, see Bauman & Lyon, 2013), and they are less concerned with subtle forms of discipline and more with brutally sorting subjects and forcing them towards ‘death worlds’ (Mbembe, 2019: 92). And who are the subjects of drone surveillance? From the warzones where killing of civilians is defined as mere ‘collateral damage’ to border policing where enslavement of migrants is defined as ‘rescue’, the subjects are those who are denied even victim status: the racialized populations of the Global South/East. To build on Daoud's observation, drones combine 17th-century racial violence with 21st-century technology. Drone technology follows in the footsteps of another major migration control technology: fingerprint biometrics. Driven by Social Darwinist racist ideology, fingerprinting was born under British colonial rule in India to identify, differentiate and control racialized populations. It was later generalized as a form of identification for the general population, though still retaining its racist core (see Pugliese, 2010). Similar patterns of racial subjugation are also observed in the political economy of drones. Frontex uses Heron and Hermes drones produced by Israeli companies and used by the Israeli military in the Occupied Palestinian Territories (MiddleEastEye, 2022), which have become a ‘laboratory for drone testing’ (Zureik, 2016: 132). These drones are later deployed in the Central Mediterranean to facilitate the racial subjugation of migrants from Africa, thus reproducing the pattern of racial (neo)colonial violence. Militarized drone surveillance of migrants has grave human rights and ethical implications. International law establishes the duty to rescue people in distress and bring them to a safe place (SAR, 1979; SOLAS, 1974; UNCLOS, 1982). Search-and-rescue (SAR) rationales are often used by authorities to frame drones as humanitarian technologies to deflect attention from their violent militaristic origins (e.g. Loukinas, 2022: 104). In practice, drone surveillance typically exposes migrants to further harm through interceptions and returns (which are strategically framed as ‘sea rescues’ – see Klein, 2021) to unsafe locations (such as Libya) rather than ensuring their safety. In fact, and as observed in the context of LCG pull-backs, drone surveillance can be the starting point for a chain of human rights violations that can even lead to the enslavement, torture, rape and death of migrants. This form of surveillance is akin to historical ‘slave patrols’ that existed to detect and return runaway slaves to captivity (see Browne, 2016: 22–24). Furthermore, the very existence of drone surveillance contributes to migrant tragedies. Similar to other border surveillance technologies, drones push migrants to find riskier routes, often at the hands of smugglers. Computerized drone vision reduces the rich complexity of migrant subjectivities, their complex life histories, ideas, identities, belongings, religious and political beliefs, hopes, ambitions, occupations, skills, pains, fears, sufferings and many other human qualities into military categories of ‘risky subjects’ or ‘threats’ from a distance. This form of dehumanization can be further explained by Bauman's concept of adiaphorization, defined as making acts, systems and processes ‘measurable against technical (purpose-oriented or procedural) but not moral values’ (Bauman, 1989: 216). The value of drone surveillance is measured by its technical capabilities (e.g. height, reach, speed, interoperability, and, more recently, AI-enabled autonomy) and its success in the objective of migration control, not by its dehumanizing effects. The physical distance between drone operators and surveilled migrants, and the data sharing with other internal or external units for intervention, further adds to the moral distancing and the abdication of responsibility for the consequences of drone surveillance. Some assassination drone operators report ‘perpetration-induced traumatic stress’ because they witness the human suffering consequences of their actions through high-resolution screens from vast distances (Pinchevski, 2016: 68). In the migration field, drone operators pass information to other units (internal or external) for intervention, and they may not witness the consequences of their actions (e.g. violence in detention). As with war drones, interpretation of the data and the decision on how to act often involves several actors, such as senior officers, legal advisors, intelligence analysts and other state officials. Intervention units might justify their actions by claiming to be ‘simply’ following orders, thus denying any personal responsibility and routinizing the dehumanization of migrants (depriving them of any moral quality), which are typical defence strategies in ‘crimes of obedience’ (Kelman & Hamilton, 1989). Drones have also been deployed in refugee camps, such as in the Greek Islands (known as ‘hotspot’ camps), with the justification that they will ‘prevent violence’ (Emmanouilidou & Fallon, 2021). The insecurity situation impacting migrants in Greek hotspots has indeed long been noted by researchers (e.g. RRE, 2018). But what causes this insecurity, and can it simply be eliminated by drones? A deeper look at the field would show that degrading camp conditions – including overcrowding, lack of hygiene, adequate food, medical care and integration support, combined with long and indefinite waits for asylum decisions – cause physical and psychological harm to migrants and constitute the root causes of insecurity in these camps (Topak, 2020). In an interview, Mustafa (pseudonym), who was living in the Moria hotspot camp of Lesvos Island shortly before it was burned down, stated, ‘Everyone is in a bad mood psychologically, and they start fighting for no reason almost’ (cited in Topak, 2020: 1869). Promoting drones as ‘technological fixes’ for insecurity in refugee camps without addressing the root causes of insecurity brings more harm to migrants by further violating their basic privacy rights and contributing to the suffering of waiting. Drones are expanding from warzones to borderzones – and further inside state territories: to refugee camps and inner cities to subjugate racialized populations. It is telling that US security officials have used drones to surveil Black Lives Matter demonstrators protesting systemic racism and police brutality (NYT, 2020). AI is increasingly integrated in border controls (e.g. Aradau, 2023; Molnar, 2023; Ozkul, 2023, in this volume), including in drones, and brings additional consequences. Frontex has several AI drone initiatives (Frontex, 2021), and Greek authorities have tested ‘autonomous drone swarms’ at the border with Turkey (Monroy, 2022). AI already ‘augment[s] the ability of drones to identify and track targets’ (Frontex, 2021: 121), and promises to make drones operate ‘fully autonomously’ even ‘in challenging environments without the need for human operators’ (Frontex, 2021: 35, 121). ‘Drone swarms’ refer to scores of autonomous drones communicating and manoeuvring with each other to achieve an objective, such as surveilling, profiling identifying, tracking and targeting ‘threats’. European drone-assisted LCG pull-backs show that drones already condemn migrants to death or near-death situations, even if indirectly. The situation is not essentially different at the Greece-Turkey borders (Evros region and Aegean Sea), where drones have been deployed and AI drone swarms are tested. These borders have long been spaces of push-backs and consequent migrant abuses and deaths facilitated by surveillance technology (Topak, 2014). AI does not alter the core violent social sorting rationale of borders/drones (see Aradau, 2023, in this volume), but rather supplements this rationale with new capabilities. AI algorithms reproduce racist patterns and categorize racialized populations as ‘high risk’ (e.g. Jefferson, 2020; see also Molnar, 2023, in this volume). Thus, algorithms can automatically categorize migrants as a ‘high risk’ (e.g. due to their movement patterns, objects they carry, or their simple presence at borderzones). This decision could be followed by human actors; or full autonomous AI drones could make automated interventions to target migrants (e.g. to prevent their passage). Both scenarios further complicate the questions of responsibility and ethics, and deepen adiaphorization and dehumanization of migrants. Yet again, the violent racializing outcomes of drones do not need them to become autonomous. Human actors interpreting drone surveillance already perceive the reality through a militarized-racialized lens, even if increasingly algorithmically mediated, and see migrants as ‘threats’; and racialized migrants already experience the brutal consequences of drone surveillance such as in Libya. Twenty-first-century drones are being used to reproduce patterns of racial violence seen in the 17th century. The dehumanizing robot eye of the drone operates remotely – not only at a physical distance from the surveilled but also at a far distance from morality. I appreciate the support of Canada's Social Sciences and Humanities Research Council (SSHRC) and research assistantship of Roberta Medina. The opinions expressed in this Commentary are those of the author and do not necessarily reflect the views of the Editors, Editorial Board, International Organization for Migration nor John Wiley & Sons.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.