Authored by Anjali Yadav, second year B.A.LLB (Hons.) student and Raja Pathak, third year student B.A.LL.B (Hons.) student at Dr. Ram Manohar Lohia National Law Iniversity, Lucknow
INTRODUCTION
Recently, in late February 2026, the United States and Israel launched air and missile strikes on Iranian nuclear sites. To counter this, Iran retaliated with missile and drone attacks against Israel and U.S. military bases in the region. The escalation has heightened regional instability and drawn international concern over a wider conflict. However, these attacks are not new or surprising in US history. The US has also previously, in June 2025, conducted an operation, Midnight Hammer, on Iran’s nuclear base, which sparked violence and retaliation by Iran at the US military base in Qatar. Amidst all of this, there is another critical issue that forms a significant part of these military operations and strikes that often goes ignored. It is the use of AI Facial Recognition Technology (“FRT”), a system that is capable of matching a human face from a digital image against a database of faces. Initially, researchers developed FRT for civilian purposes such as security and verifying identities at checkpoints. However, oHwith the passage of time, this technology has gained momentum and have used in several substantial tasks as well, but at the same time, it has sparked debate on its potential misuse.
This blog critically examines the legal, ethical and humanitarian challenges arising from the use of FRT in armed conflict, specifically its implications for International Humanitarian Law and human rights. It further explores the risks of bias, misuse and lack of accountability associated with such technologies. The discussion then presents plausible, effective solutions, including purpose-bound regulation, enhanced safeguards and stronger enforcement mechanisms to ensure lawful deployment and statutory recognition in line with evolving technological realities.
Rising Legal And Human Rights Concerns In Facial Recognition
Currently, the prevalence of facial recognition technology systems has increased significantly, and organisations use them for a multitude of purposes, benefiting people in several aspects, such as enhanced security, fraud prevention, and modern warfare. However, while it has aided in improving precision in identifying insurgents, it also poses a serious legal and ethical risk, which indicates a foreshadowing of potential adverse events.
A. Violation of International Humanitarian Law (IHL) Principles
The principles of distinction, proportionality, and precaution are absolute in each of the Geneva Conventions and Additional Protocols. The principle of distinction requires parties to an armed conflict to clearly distinguish between combatants and civilians and direct their operations only against legitimate military targets. However, the documented flaws of FRT, particularly in relation to skin tones and genders, explicitly undermine this principle, as they increase the risk of misidentification and compromise the obligation of belligerents to accurately differentiate between combatants and civilians.
A study published in 2024 by the International Review of the Red Cross considers the implications of these technologies and warns that “the reliance on machine-classified identity without human corroboration is inconsistent with IHL obligations.” Hence, misidentifications may lead to unlawful or indiscriminate attacks.
B. Failures Under International Human Rights Law (IHRL)
Article 17 of the International Covenant on Civil and Political Rights (“ICCPR”) seeks to protect the privacy rights or deter unlawful interference with someone’s reputation even during the state of armed conflict, but it is often violated by mass FRT surveillance. For example,Israel’s Blue Wolf and Red Wolf systems monitored and scanned Palestinians through extensive biometric databases. Despite concerns raised by several human rights organisations that algorithms and AI technologies may reinforce racial or social inequalities by relying on biased biometric data and design frameworks, commonly known as “algorithmic apartheid”, the lack of effective enforcement mechanisms continues to erode the rights to privacy and human dignity.
Importantly, the UN Guiding Principles on Business and Human Rights (UNGPs) framework creates an obligation for the private FRT vendors to prevent this harm under the second pillar of UN Guiding Principles on Business and Human Rights (UNGPs), which particularly highlights the corporate responsibility to respect human rights. But on the other hand, military contracts rarely offer robust data protections or deletion obligations, thus exposing civilians to wrongful or lasting misuse.
C. Discriminatory Bias and Algorithmic Opacity
Scholars and Studies have well documented that FRT is applied in inconsistent ways to different demographics. For instance, the U.S. National Institute of Standards and Technology (NIST) published a study in 2019 to weigh the performance of algorithms from the U.S. It found that algorithms consistently showed a high racial disparity. The error rates for African American and Asian faces were approximately 10 to 100x higher than the white faces. Similarly, the “Gender Shades” project by the MIT Media Lab found an error rate of just 0.8% for light-skinned men, but an error rate as high as 34.7% for dark-skinned women. In mixed-ethnicity conflict zones, the risk of introducing ethical and legal bias could exacerbate ethnic targeting or collective punishment that violates Articles 2 and 26 of the ICCPR on non-discrimination. In addition, deep-learning-based FRT often operates as a “black box,” which makes it hard for commanders to verify the system’s logic or ensure compliance with the Article 144 of Geneva Convention (IV) relative to the Protection of Civilian Persons in Times of War 1949 (“Geneva Convention IV”), which requires states to educate armed forces on lawful targeting methods.
D. Automation Bias and Overdependence
Military operators are often susceptible to automation bias, thereby holding excessive faith in algorithmic decisions, which leads them to overlook evidence to the contrary. In the case of Williams v. City of Detroit, where Detroit police wronglyarrested a man in a matter pertaining to burglarysolely based on the use of FRT.The court further held strict, federal court-enforced policies against FRT for four years, including training given to police officers on FRT dangers, requiring independent evidence beyond a match, and auditing past FRT cases.Similarly, in Bridges v South Wales Police, the UK Court of Appeal held that the deployment of FRT by South Wales Police was unlawful and a violation of Article 8 of the European Convention on Human Rights, as well as failures to comply with data protection and equality obligations. On a parallel note, in a class-action suit against Macy’s in Illinois, the plaintiff accused of collecting biometric data without consent, which has sparked an emerging trend of holding private entities accountable for intrusive surveillance practices.
Furthermore, where the accuracy of FRT is in question, several military organisations are also involved in “Terror Capitalism”, a process that combines economic demands for data and cheap labour with state incentives to develop and test digital and automated surveillance in the name of security.This not only cites concerns for data violation but also stands contrary to Article 57 of Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to the Protection of Victims of International Armed Conflicts (Protocol 1) (“Additional Protocol I”), which requires that parties make appropriate, diligent efforts to verify that targets are lawful military objectives.
Way Forward
To reconcile FRT with humanitarian and human rights norms, there is a need for a comprehensive solution and required laws for FRT. Some of the solutions are stated below.
A. Legally Mandated Purpose-Bound and Time-Limited FRT Use for Identifying the Dead
All States should establish a time-bound and purpose-restricted system that permits the use of FRT or biometric data exclusively for the identification of deceased soldiers or civilians, and ensure that authorities must strictly regulate such deployment. This obligation must operate in full compliance with Articles 15 to 17 of the Geneva Convention (I) for the Amelioration of the Condition of the Wounded and Sick in Armed Forces in the Field (“Geneva Convention I”), which collectively governs the treatment and identification of the wounded and the dead in armed conflict. In particular, Article 15 specifies that parties search for, collect and evacuate the wounded and sick, while also protecting the dead from despoilment or neglect. Article 16 requires recording and transmission of all available information such as identity discs, documents and personal effects, to facilitate identification and notification of families. Furthermore, Article 17 obligates States to ensure the honourable burial of the dead, proper maintenance and marking of graves, and the adoption of measures that enable future identification.
A time-bound and purpose-restricted system must lock data by design for use strictly for repatriation and notification of families, and must prohibit any form of intelligence-gathering or profiling. The system must automatically erase the data within a specified period once the identification process is complete.
The system should also incorporate independent verification by the International Committee of the Red Cross (ICRC) or other neutral agencies to certify secure deletion. It can achieve this through widely available tamper-evident systems that log data access and provide real-time alerts of any unauthorised queries. These safeguards will uphold the dignity of the deceased in accordance with International Humanitarian Law, prevent misuse by traffickers or other malicious actors, and provide clarity and operational feasibility for military authorities.
B. Multi-Source Validation and Bias-Guard Rule
To reduce discriminatory bias, authorities must ensure that no FRT match alone justifies targeting or detaining a person. They should validate such matches using at least two independent intelligence sources. This safeguard must conform to international legal obligations, particularly under Articles 2 and 26 of ICCPR, which not only require States to respect and ensure all Covenant rights without distinction of any kind but also impose a positive obligation to prevent discriminatory practices and provide equal and effective protection of the law against any form of discrimination. Similarly, Article 57 of the Additional Protocol I to the Geneva Convention establishes the principle of precaution in attacks, mandating that parties must take all feasible steps to verify that targets are lawful military objectives, choose means and methods that minimize incidental civilian harm. It should refrain from action where there is doubt.
Authorities must also provide troops with legal training on biases inherent in FRT systems and deploy systems equipped with explainability modules that allow commanders to review and assess the basis of automated decisions.
Furthermore, in order to curb “Terror Capitalism”, authorities must prohibit the use of civilian data collected under security pretexts for commercial or exploitative technological development, and must ensure adequate legal protections to safeguard human dignity.
CONCLUSION
As warfare moves from the battlefield to cyberspace, FRT will emerge as an invisible weapon that determines who dies, who lives, and who is wrongly labelled a threat. While governments may present FRT as a precision tool, flawed algorithms and opaque surveillance technologies risk turning war zones into laboratories for human rights violations.
To prevent technology from undermining humanity, states must urgently adapt proper enforcement mechanisms by incorporating time-bound and purpose-restricted systems and restricting discriminatory bias of all kinds. Purposeful and time-bound data storage, multi-layers of verification, and independent oversight are not options but are necessary to protect human dignity amid conflict. Thus, if the world chooses to automate without accountability, technology will weaponise injustice rather than protect human life. Technology should serve humanity, not enslave humanity.
Leave a Reply