Introduction
For more than 160 years, the Red Cross has symbolised something almost sacred on the battlefield: the idea that certain spaces and people are protected from conflict and harm. Delegates of the Red Cross were trusted as politically neutral actors whose sole objective was to alleviate human suffering. Hence, they were able to enter battlefields unarmed, cross front lines, visit prisoners of war, and negotiate with opposite sides.
This neutrality has never been merely an altruistic ideal; it has been the foundation that allows the International Committee of the Red Cross(ICRC) to gain access to victims regardless of their nationality, ideology, or military allegiance. Yet the conditions that once allowed this principle are changing rapidly. Due to rapidly evolving technology, modern warfare is increasingly shaped by cyber operations, hybrid conflicts, non-state armed groups, and, most significantly, artificial intelligence(AI)—all of which complicate traditional distinctions that created the conditions under which neutrality worked effectively. Although the ICRC continues to uphold neutrality as one of its fundamental principles, maintaining operational neutrality has become increasingly difficult. This paper will argue that, while the notion of neutrality will remain valid for the ICRC, it will require a significant overhaul to sustain its diplomatic leverage in an era of increasingly AI-driven warfare, albeit in a much more limited capacity than in the past.
Why Has Operational Neutrality Worked Historically?
Historically, due to the structure of warfare, humanitarian mediation was supported naturally, leading to the success of operational neutrality. Conventional interstate wars generally involved sovereign states with clearly defined borders, recognised governments, identifiable armed forces, and formal chains of command. Combatants wore uniforms, battlefields occupied physical locations, and responsibility for military actions could usually be traced to specific authorities. These conditions made it comparatively straightforward to distinguish civilians from combatants, humanitarian space from military objectives, and lawful military targets from protected persons. More importantly, governments had strong incentives to respect humanitarian neutrality because doing so ultimately protected their own soldiers as much as their enemies’.
This mutual interest is manifested by the remarkable acceptance of the 1864 Geneva Convention that was ratified by all major European powers within only three years before eventually forming the basis of the 1949 Geneva Conventions that are accepted by more than 180 states(International Committee of the Red Cross [ICRC], 2013). Such rapid ratification was not simply an expression of humanitarian goodwill; rather, it reflected widespread recognition that neutral medical personnel and humanitarian organisations benefited every participant in armed conflict. In this context, the alignment of moral principles with the states’ long-term strategic interests led to the acceptance of operational neutrality.
This led to the construction of legal frameworks that enabled humanitarian access in the ICRC’s later work, such as restoring contact between POWs and their families during the First World War(ICRC, 2015) and negotiating access to approximately 150,000 North Korean and Chinese prisoners held by the United Nations Command during the Korean War(Humanity Journal, 2023). These were all possible because identifiable governments retained the authority to negotiate, recognise neutral actors, and enforce international humanitarian law. Consequently, neutrality earned trust not merely because the ICRC declared itself impartial, but because the structure of conventional warfare provided actors capable of recognising and respecting that impartiality.
The Transition: The Nature of Conflict Changed
Though the ICRC had maintained neutrality throughout, those conditions that allowed the neutrality began to erode long before AI entered the battlefield. The shift towards warfare such as terrorism, cyber warfare, hybrid operations, and non-state armed groups arose in the years after the Cold War, and particularly after the attacks of 11 September 2001(Robertson, 2025). These forms of conflict rarely possess clearly defined front lines, universally recognised combatants, or governments exercising complete control over every participant, making them very different from conventional state wars.
One of the early indications of this transition emerged during the United States’ War on Terror, well before cyber operations entered the picture. Detainees captured in Afghanistan and elsewhere and held at Guantanamo Bay were classified not as prisoners of war under the Third Geneva Convention, but as “unlawful” or “enemy” combatants, a designation the ICRC challenged because it bypassed the legal protections on which humanitarian neutrality depended(ICRC, n.d.). If states could redefine who qualified for protection under international humanitarian law, the legal categories supporting operational neutrality would become significantly less stable.
In addition to cyberwarfare, cyberattacks have added speed to this shift. The Stuxnet 2010 cyberattack against Iran’s nuclear facilities is an example of the first attack to take place on state infrastructure and which is attributed to a joint American-Israeli mission(Fruhlinger, 2022). This attack marked the appearance of a new grey zone between peace and war. However, the 2017 NotPetya attack demonstrates clearly why the issue of operational neutrality became even more challenging. The attack, initiated by the Russia-based group of hackers called Sandworm in the midst of the ongoing Russia-Ukraine conflict, quickly became uncontrollable. NotPetya resulted in a financial loss to Ukraine equal to about 0.5% of GDP of the country(Johansmeyer, 2025) while also damaging vaccine manufacturing for the pharmaceutical firm Merck(InfoSec Labs, 2026).
The significance of NotPetya extends beyond its scale. Traditionally, the ICRC secured humanitarian access by entering identifiable battlefields and negotiating with responsible military authorities. Cyber attacks such as NotPetya possess neither. Propagating through interconnected digital infrastructure, the malicious code released affects networks far removed from the original conflict, ranging from business and healthcare-related systems to civilian networks. Thus, there is often no clear frontline to cross, no single authority capable of containing the attack, and no geographically defined humanitarian space in which neutrality can operate. Rather than simply introducing new weapons, cyber warfare fundamentally altered the environment in which humanitarian neutrality had previously functioned, marking the first major challenge to the ICRC’s operational model.
AI Warfare: The Next Wave of Challenges
If cyber warfare had weakened the foundations of operational neutrality, artificial intelligence threatens to transform them even more fundamentally by changing who or what makes military decisions. The increasing use of AI for analysis, surveillance, and target selection reduces meaningful human judgement within military operations. The clearest example is Israel’s reported use of AI systems Lavender and The Gospel during the ongoing Gaza conflict. According to an investigation by +972 Magazine, these systems generated approximately 37,000 potential targets, while officers reportedly spent as little as twenty seconds reviewing each AI-generated recommendation before approving strikes(Abraham, 2024). The significance of this evidence lies not only in the speed of decision-making but in what it suggests about human agency.
Although humans technically remained during the process, a twenty-second review scarcely constitutes independent military judgement. The question of responsibility becomes all the more complicated because until now, responsibility for military decisions always lay on the person who issued that particular order or approved the use of force. This shift complicates accountability because responsibility for military decisions has traditionally rested with the individual issuing an order or authorising the use of force. When AI increasingly shapes those decisions, it becomes unclear whether responsibility lies with the commander, the software developers, the military institution, or the algorithm itself.
The growing integration of commercial AI into military operations further complicates this problem. In July 2025, Anthropic signed a US$200 million contract with the United States Department of Defence, and Claude was reportedly used through Anthropic’s partnership with Palantir during the January 2026 operation that captured Venezuelan leader Nicolás Maduro(Built In, 2026). Shortly afterwards, Defence Secretary Pete Hegseth demanded that Anthropic remove restrictions on autonomous weapons and domestic surveillance(Philips, 2026). Although the company refused(Haynes, 2026), President Trump subsequently ordered federal agencies to cut ties with Anthropic, which led the Pentagon simply to shift to alternative providers(MEXC News, 2026).
This reveals that decision-making authority is increasingly distributed not only across governments but also among private technology companies whose international policies influence military capabilities, and that the ICRC’s traditional model of neutrality faces an unprecedented challenge. The historical importance of neutrality, however, lied in the possibility of negotiating with identifiable decision-makers who were capable of weighing the needs of warfare with their humanitarian duties. But with the increasing involvement of algorithms in making target decisions and the role that corporations play in developing military AI, it is difficult to identify those who hold any significant amount of power. It is important, thus, not only to ask about the neutrality of the ICRC, but also about the value of this neutrality amid new actors in the conflict.
How the Red Cross Is Responding
Despite these challenges, it would be inaccurate to argue that the ICRC has remained passive. Rather than abandoning neutrality, the organisation has sought to reinterpret how it should operate within an increasingly digital battlespace. In May 2025, ICRC president Mirjana Spolijaric renewed the organisation’s October 2023 joint appeal(ICRC, 2023) with the United Nations Secretary-General for a legally binding international treaty regulating autonomous weapons systems, arguing that meaningful human control must remain central to decisions over the use of force(ICRC, 2025). This demonstrates an important change in strategy by the ICRC. Neutrality used to be concerned mostly with ensuring humanitarian access once conflict had broken out.
Today, however, the organisation is trying more often to influence the rules regulating the development of new military technologies before their integration into warfare. In addition, the ICRC has initiated the Digital Emblem Project, whose goal is to develop an emblem for cyberspace akin to the Red Cross emblem to mark humanitarian entities during cyber operations (ICRC, 2026a). Just as the original emblem communicates protected status on physical battlefields, the digital emblem seeks to establish similar protection within digital networks where traditional humanitarian symbols have little practical meaning. Additionally, the Seventh Review Conference of the CCW was planned by the ICRC in November 2026 as an initial step towards negotiating on autonomous weapons (ICRC, 2026b). The above two moves indicate that the ICRC is shifting its interpretation to align with the challenges faced by modern-day warfare without compromising its principles of neutrality.
The Limits of These Responses
Though the initiatives above show that the ICRC is actively adapting, they do not fully address the challenges arising from AI-driven warfare. The first problem is speed. AI technology is advancing far faster than humanitarian law can respond. Claude had already been integrated into classified Pentagon networks by late 2024 and received a contract with the Department of Defence by July 2025(The Neuron, 2026), yet despite having more than 120 states supporting the negotiation on autonomous weapons(Human Rights Watch, 2025), no formal negotiations had begun, with the next major opportunity not scheduled until the CCW Review Conference in November 2026(ICRC, 2026b). By the time governments begin discussing regulation, the technology has often already been deployed.
At the same time, decision-making is no longer controlled solely by states; private companies and multinational tech firms now hold real influence, as shown when Anthropic refused to remove restrictions on autonomous weapons for the Pentagon due to its internal policy(Haynes, 2026). Since ICRC’s model assumes states and armed groups as the main actors, such a change in decision-making power, now sitting partly within corporate structures, makes it inapplicable. Economic incentives further complicate regulation because when Anthropic refused to loosen its restrictions, the Pentagon did not stop but instead signed agreements with other companies, including OpenAI, Google, Microsoft, Nvidia, AWS, Oracle, SpaceX, and Reflection(Built In, 2026), showing that no single company can act as a barrier and that competitors are financially motivated to fill any gap, allowing regulation aimed at one firm to be bypassed through market competition.
AI deployment consistently outpaces the slow, consensus-driven process of international regulation, leaving humanitarian law perpetually reacting to technologies already in use. At the same time, the rise of private tech companies as independent decision-makers undermines the ICRC’s state-centric model, since market competition allows firms to bypass restrictions that any single company refuses to accept. Taken together, these dynamics show that the ICRC is trying to operate within a system that is changing too quickly, making it harder to maintain neutrality—not because the ICRC lacks commitment, but because the conditions that once supported neutrality are no longer stable.
Rethinking Humanitarianism
If AI-driven warfare has fundamentally changed the conditions under which neutrality can function, then the challenge no longer lies in whether the ICRC should remain neutral, but whether neutrality alone is sufficient. Traditionally, organisations generally focused on humanitarian work during conflict, yet AI shapes humanitarian outcomes long before the first humanitarian worker enters a battlefield. Thus, the ICRC’s response by trying to legally bind restrictions on autonomous weapons and to develop initiatives represents recognition of the idea that protecting civilians requires more than responding to war.
This, however, is not abandoning neutrality. Instead, as a response to this evolving AI- driven world, neutrality should evolve alongside warfare, combining impartial humanitarian action during conflicts with greater involvement in shaping international norms before conflicts begin. Remaining neutral may therefore require becoming more proactive rather than simply remaining reactionary.
Conclusion
In conclusion, the notion of neutrality will remain valid for the ICRC as a humanitarian principle, but sustaining it as an operational neutrality will require a significant overhaul, as conditions that once made it effective no longer exist in the same form. As warfare becomes progressively shaped by AI, the organisation may still retain some influence over how conflicts are conducted. Still, that influence is likely to be far narrower than it was. Historically, ICRC’s neutrality principle was effective because it was a part of a system where the nature of war itself sustained humanitarian mediation almost automatically, allowing operational neutrality to function without significant strain. However, modern warfare’s landscape has steadily dismantled these conditions.
Cyber attacks have demonstrated that conflict is no longer confined to battlefields, and AI-driven warfare systems are increasingly entangling military decisions with opaque algorithms. Furthermore, as competition among private technology companies and complex geopolitical tensions increase, human agency in such conflicts is severely limited. Although the ICRC has responded by promoting international regulation and adapting humanitarian practice to the digital age, these measures remain constrained by the speed of technological change and its growing influence. Neutrality will remain the moral foundation of any humanitarian action, but whether it can also sustain its operational foundation is far less certain. As artificial intelligence continues to transform the nature of warfare, humanitarianism itself may need to evolve in ways that previous generations of humanitarian law never anticipated.
Bibliography
- Abraham, Y. (2024, April 3). ‘Lavender’: The AI machine directing Israel’s bombing spree in Gaza. +972 Magazine. https://www.972mag.com/lavender-ai-israeli-army-gaza/
- Built In. (2026, February). Anthropic vs. Pentagon: Fight over Claude access. https://builtin.com/articles/anthropic-pentagon-claude-dispute
- Fruhlinger, J. (2022, August 31). Stuxnet explained: The first known cyberweapon. CSO Online. https://www.csoonline.com/article/562691/stuxnet-explained-the-first-known-cyberweapon.html
- Haynes, K. (2026, February 16). Anthropic blasts Pentagon’s use of its AI tool in Venezuela raid—May void $200 million contract. Inc. https://www.inc.com/kevin-haynes/anthropic-blasts-pentagons-use-of-its-ai-tool-in-venezuela-raid-may-void-200m-contract/91303320
- Human Rights Watch. (2025, May 21). UN: Start talks on treaty to ban ‘killer robots’. https://www.hrw.org/news/2025/05/21/un-start-talks-treaty-ban-killer-robots
- Humanity Journal. (2023). The major humanitarian dilemma of neutrality: The International Committee of the Red Cross and prisoners of war in Korea, 1950–1953. https://humanityjournal.org/issue13-3/the-major-humanitarian-dilemma-of-neutrality-the-international-committee-of-the-red-cross-and-prisoners-of-war-in-korea-1950-1953/
- InfoSec Labs. (2026, April 23). NotPetya cyberattack analysis: Impact & mitigation strategies. https://infoseclabs.io/blog/decoding-notpetya-the-1-4-billion-cyber-catastrophe-and-its-lessons
- International Committee of the Red Cross. (2013, January 25). The 1864 Geneva Convention. https://www.icrc.org/en/doc/who-we-are/history/150-years/timeline/1864-convention.htm
- International Committee of the Red Cross. (2015, January 29). The International Committee of the Red Cross in the First World War. https://www.icrc.org/en/document/international-committee-red-cross-first-world-war-0
- International Committee of the Red Cross. (2023, October). Joint call by the United Nations Secretary-General and the President of the International Committee of the Red Cross for States to establish new prohibitions and restrictions on Autonomous Weapon Systems. https://www.icrc.org/en/document/joint-call-un-and-icrc-establish-prohibitions-and-restrictions-autonomous-weapons-systems
- International Committee of the Red Cross. (2025, May 12). Preserving human control over the use of force: A call to regulate lethal autonomous weapon systems under international law. https://www.icrc.org/en/statement/preserving-human-control-over-use-force-call-regulate-lethal-autonomous-weapon-systems
- International Committee of the Red Cross. (2026a, January 15). IHL’s lighthouse: Navigating towards a digital emblem. https://blogs.icrc.org/law-and-policy/2026/01/15/ihls-lighthouse-navigating-towards-a-digital-emblem/
- International Committee of the Red Cross. (2026b). Advocacy paper: A key opportunity to prevent the development of unacceptable autonomous weapons. https://www.icrc.org/en/article/advocacy-paper-key-opportunity-prevent-development-unacceptable-autonomous-weapons
- International Committee of the Red Cross. (n.d.). Unlawful combatants. In ICRC casebook: How does law protect in war? https://casebook.icrc.org/a_to_z/glossary/unlawful-combatants
- Johansmeyer, T. (2025, October 15). NotPetya, Ukraine, and the limits of economic impact from cyber attacks. The Loop, ECPR. https://theloop.ecpr.eu/notpetya-ukraine-and-the-limits-of-economic-impact-from-cyber-attacks/
- MEXC News. (2026, March 5). Anthropic’s $200M Pentagon deal falls apart — Now negotiations are back on. https://www.mexc.com/news/859443
- Phillips, M. (2026, February 24). Pentagon gives AI firm ultimatum: Lift military limits by Friday or lose $200M deal. Fox News. https://www.foxnews.com/politics/pentagon-gives-ai-firm-ultimatum-lift-military-limits-friday-lose-200m-deal
- Robertson, A. (2025). Editorial – Transformation of war. Journal of Military and Veterans’ Health, 33(1). https://jmvh.org/article/editorial-34/
- The Neuron. (2026, March 2). Anthropic vs the Pentagon vs OpenAI: The full story. https://www.theneuron.ai/explainer-articles/the-pentagon-vs-anthropic-explained-the-week-ai-drew-a-line-in-the-sand-and-the-government-kicked-it-over/
