Why we must keep humans at the heart of AI in warfare

The Herald Scotland

16 hours ago

Since 2016, discussions of the Convention on Certain Conventional Weapons Group of Governmental Experts on LAWS have been ongoing, but International Humanitarian Law (IHL) still lacks any specific, binding regulations relating to AI. As noted by International Committee of the Red Cross (ICRC) President Mirjana Spoljaric, AI in war is 'no longer an issue for tomorrow', but rather 'an urgent humanitarian priority today', requiring the immediate 'negotiation of new legally binding international rules'. Accordingly, United Nations Secretary General António Guterres recommended, in his 2023 New Agenda for Peace, that 'a legally binding instrument' to prohibit and/or regulate AI weapons be concluded by 2026.
Read more
The ICRC has stressed that responsibility in warfare must remain with humans. 'Human control must be maintained,' it argues, and limits on autonomy urgently established 'to ensure compliance with international law and to satisfy ethical concerns'.
In 2022, the MoD itself echoed this sentiment. It stated that only human soldiers 'can make instinctive decisions on the ground in a conflict zone; improvise on rescue missions during natural disasters; or offer empathy and sympathy.' The then Defence Secretary Ben Wallace added that 'at its heart, our Army relies on the judgment of its own individuals.'
A recruitment campaign at the time carried the tagline: 'Technology will help us do incredible things. But nothing can do what a soldier can do.'
Colonel Nick Mackenzie, then Assistant Director for Recruitment, highlighted that, while 'technology is really, really important… there is always somebody, a person, behind that technology,' who is ultimately responsible for its use and the decisions it enables.
Since then, however, the use of AI-enabled rapid target identification systems in contemporary conflicts has grown rapidly, with notable examples being Lavender and Where's Daddy (Israel/Palestine), Saker and Wolly (Russia/Ukraine). A human being is generally still required in order to engage any lethal effects, but technological capabilities are already being developed to remove human input from the targeting process altogether.
Against this backdrop, the MoD's Strategic Defence Review 2025, released last month, calls for 'greater use of autonomy and Artificial Intelligence within the UK's conventional forces' to deliver 'greater accuracy, lethality, and cheaper capabilities'. 'As in Ukraine,' the Review continues, 'this would provide greater accuracy, lethality, and cheaper capabilities – changing the economics of defence.'
One example is Project ASGARD, which will help the Army locate and strike enemy targets at greater distances using AI as a 'force multiplier'. This is just one of over 400 AI-related projects being run by the MoD.
What remains unclear, but is critical from a legal and moral perspective, is what role human judgment will play in these projects and the military operations they support.
Computer scientist Pei Wang has said that while AI can behave like human intelligence in some ways, it is fundamentally different. AI shouldn't replace human intelligence, but rather support and enhance it – helping people make better-informed decisions.
Human-robot interaction specialist Karolina Zawieska warns of the need to distinguish between what is human and what is only human-like. AI systems often function as a 'black box', meaning it is not always clear how or why they produce certain outcomes. This creates serious problems for human understanding, control, and accountability.
When properly used, AI can support situational awareness and help human operators make better decisions. In this sense, it is a tool – not a decision-maker. But if too much control is handed over to AI, we risk removing human judgment and with it, moral responsibility.
Professor Jeff McMahan, moral philosopher at the Oxford Institute for Ethics, Law and Armed Conflict, has argued that it is essential for combatants to feel 'deep inhibitions about tackling non-combatants'.
However accurate or efficient AI may be, these inhibitions cannot be replicated by algorithms. As political scientist Valerie Morkevičius has pointed out, the emotional and moral 'messiness' of war is a feature, not a flaw because it slows down violence and prompts ethical reflection. Military decisions should be difficult. This is why human judgment must remain at the centre.
While defence and national security are reserved for Westminster, Scotland plays a key role in UK defence, from the bases at Faslane and Lossiemouth to the defence research carried out at Scottish universities. The issues raised in the Strategic Defence Review therefore carry particular relevance here.
UN Secretary General António Guterres has recommended that 'a legally binding instrument' to prohibit and/or regulate AI weapons be concluded by 2026 (Image: Getty)
Scotland's approach to AI, shaped by the AI Strategy (2021) and the Scottish AI Playbook (2024), is notably human-centred. Informed by Organisation for Economic Cooperation and Development's (OECD) principles, both documents stress the importance of trustworthy, ethical, and inclusive AI that improves people's lives. They highlight the need for transparency, human control, and robust accountability.
Though not military in scope, these principles nevertheless offer a useful framework for a Scottish perspective on the development and use of AI for military purposes: keeping people at the centre, and ensuring that technology supports rather than replaces human agency.
The goal should not be the delegation of human decisions to machines, or the replacement of human beings with technology. Rather, AI should support and strengthen human decision-making – a tool for the enactment of human agency: a technological means for strictly human ends.
Dr Joanna LD Wilson is a Lecturer in Law at the University of the West of Scotland