This article is an opinion piece whose contents represent the standpoint of its author and not UPF Lund or The Perspective’s editorial board.
Artificial Intelligence (AI) technologies do not exist without human influence; there is always a risk of prejudices, biases, and harmful beliefs seeping their way into algorithmic infrastructures. When integrated into military operations, these technologies can influence decisions that determine whose lives are put at risk.
Thus, military AI programs cannot exist without the impacts of their creators, which can reproduce existing inequalities. The Stockholm International Peace Research Institute identifies three sources of bias in military AI: bias in society, bias in data processing and algorithmic development, and bias in use. Various militaries already employ AI technologies, such as the use of programs that aid workflows or target software.
When existing assumptions are translated into data-driven classifications, the consequences extend beyond technical inaccuracies or algorithmic errors, potentially to the acceleration of violence against targeted populations.
Israel: Algorithmic Targeting
The Israel Defense Forces (IDF) have developed an advanced AI targeting system known as Habsora (‘The Gospel’), which has reportedly been in operational use since 2021. It processes large volumes of data across multiple sources to identify ‘physical objects’ that are potentially affiliated with labeled terror groups. Intelligence analysts then review these recommendations and make judgements on whether the output is accurate.

A separate Israeli AI system, Lavender, received great attention internationally in April 2024 as +972 Magazine and Local Call published an investigation into the use of the technology. The two outlets reported that Lavender generated potential bombing target lists of suspected individuals affiliated with Hamas and the Palestinian Islamic Jihad.
The authors call on firsthand accounts from IDF intelligence officers, indicating that Lavender had limited human verification, often with personnel serving as ‘rubber stamps’ and conducting brief checks, such as simply determining whether a target was male.
The IDF has since described these claims as false.
This raises broader questions about the algorithms themselves. The problem extends beyond misidentification; these AI systems may use patterns or associations that can then lead to processes involving the use of lethal force taken against humans.
United States: Expansion of Commercial AI
Across the Atlantic, in the United States, there is a growing relationship between commercial AI companies and the military. In May 2026, the United States Department of War announced agreements with eight leading AI companies, including OpenAI, SpaceX, Google, and Microsoft, to deploy their technologies for military work.
Additionally, the Department has developed an official internal AI platform, GenAI.mil, which is reportedly used by over 1.3 million civilian and military personnel to support administrative tasks.

Following this, in August 2026, the Department of War announced ChatGPT Mil, a new step for the department’s internal platform. This progression is in line with the White House’s AI Action Plan and seeks to further accelerate workflows, allowing personnel to focus their attention on projects deemed more critical.
Although these AI applications are not targeting systems like Lavender, their adoption showcases how the technology is becoming more embedded into the administrative infrastructure of the military. As these technologies are used to support planning, research, and operational work, this raises questions of whether existing assumptions and biases are being actively reproduced in these automated processes.
Europe: AI-Controlled Defense Systems
Helsing, a German defense technology company founded in 2021, has recently unveiled Contra, a new AI-controlled air defense and threat coordination system. Designed to detect and engage large numbers of incoming drones and missiles, Contra illustrates a different application of AI compared to systems that classify potential targets or increasing workflows. The company already holds several contracts across Europe, including in the United Kingdom, Sweden, Ukraine, and Germany.
The use of Contra , however, raises questions about the consequences of incorrectly identified threats, and the accountability of those involved in acting on automated assessments.
Acceleration of Violence and Displacement of Responsibility
Language used by military AI operations often indicates the prioritization of efficiency, productivity, and technological precision. As the speed of intelligence processing and potential target identification increases, questions are raised about the consequences of these AI systems. When recommendations are generated at a rate that greatly exceeds human capabilities, is there a risk of less human oversight if recommendations are routinely accepted or validated?
When an AI system incorrectly identifies information, who is ultimately held responsible when the decisions result in wrongful targeting or civilian death? Is it the tech developers, the personnel interpreting the outputs, the commanders authorizing the military action? Will this decentralization of power lead to a diffusion of responsibility? Further, how might integration of AI into military operations shape civilian populations through increased surveillance, misidentification risks, and the erosion of existing protections?
AI Security and Defense Governance
As various applications of AI continue to expand in military operations, international discussions increasingly address how states should govern its development. For example, the United Nations Institute for Disarmament Research established its Centre of Excellence on AI, Peace and Security in June 2026 to tackle the lack of continuity between short-lived policy processes.
The center recently held a forum, bringing together various stakeholders to discuss the consequences of AI-based defense and security. In addition, the center published guidelines for national strategies on AI-related security and defense measures in October 2026.
These initiatives reflect growing institutional attention to state-used AI in security and defense operations. Ultimately, the implications of military AI extend beyond whether systems can accurately identify threats or operate more efficiently than humans. Governance must also address how these AI systems define a threat; who becomes subject to that classification, and what happens when violence is facilitated by bias?
By: Jocelyn Pembleton
November 17, 2026








