At UN talks on military AI, Stop Killer Robots calls for strong rules against the automation of violence

Last week, between 15-17 June 2026, states met at the UN in Geneva to discuss their views on how to respond internationally to the issue of ‘AI in the Military Domain.’ These international discussions were mandated by United Nations General Assembly resolution 80/58,  and came in the context of increasing reports of harm from AI in warfare, as well as the growing pursuit of AI and automation in the use of force by states and industry for its possible ‘benefits’. Stop Killer Robots joined the talks to highlight the huge risks and harm being caused by this trend – and to urge states to start taking immediate measures and agree to concrete international rules to prevent these now.

As a global coalition of more than 300 member organisations in over 75 countries, Stop Killer Robots campaigns to ensure meaningful human control over the use of force, resist digital dehumanisation, and prevent automated harm. Marginalising meaningful human control and judgement in favour of an overreliance on artificial intelligence (AI) and automation dehumanises and accelerates violence, which threatens civilian protection, risks violations and challenges to accountability, and also risks the escalation and intensification of conflicts.

Evidence is mounting of the use of AI by militaries, as are reports of its link to serious harm to civilians. Israel’s use of AI tools to suggest human and other targets in Gaza has contributed to digital dehumanisation and widespread humanitarian consequences. AI’s role in enabling conflict parties to accelerate the pace of war—which risks increased errors and harm—has also been prominent in recent operations by the US and Israel in Iran and Lebanon. Various digital tools to support operations are being used by Russia and Ukraine in the conflict in Ukraine, and are being pursued by a wide range of countries.

Since 2013, Stop Killer Robots has been campaigning for states to negotiate a treaty on autonomous weapons systems (AWS)—but with the changing technological and political landscape we are witnessing, we believe the international community should also support the development of further norms to address other aspects of and concerns with AI and automation in the use of force across armed conflict and law enforcement. International discussions on autonomous weapons have thus far excluded these matters from consideration, while those on “military AI” have so far only produced principles and declarations focusing on “responsible use” that have taken a permissive approach. A stronger international response is needed. 

States’ discussions on how to regulate AWS have taken place predominately at the Convention on Certain Conventional Weapons (CCW), but have also advanced through various forums and processes including through the United Nations General Assembly (UNGA) and regional and international conferences, and are now at a stage where negotiations on an instrument can and must start. Through the constructive work of states under the current mandate of the Group of Governmental Experts (GGE) on lethal autonomous weapons systems (LAWS), a “rolling text” of possible elements of an instrument on autonomous weapons systems has been developed that can provide a good basis for negotiations, and over 70 states have said they are willing to move forward on this basis. While critical issues are missing from current discussions, foremost the need to prohibit anti-personnel AWS, states are now faced with a huge opportunity to start negotiations on a legal instrument, which could set clear and meaningful lines on keeping meaningful human control in the age of AI and automation.

But the international community cannot stop at agreeing on a treaty on autonomous weapons alone. When it comes to AI in the use of military force beyond the specific boundary of AWS, two current issues raise particular concerns for us. These are: 1) The use of AI and automation in the data systems of the so-called kill chain of identifying and attacking targets, including systems that identify and recommend targets, and those that accelerate the production of target recommendations based on data accumulated through mass surveillance, or that integrate generative AI systems; and 2) other efforts to use AI and automation for “force multiplication” in ways that test the cognitive limits of people to use systems—for example the potential for one person to be charged with the control of multiple drone swarms. The conversation on AI in the military domain is wide ranging, but for us these are the two key issue areas where the risks of harm are greatest; and in the case of AI in the kill chain, we have already seen clear reports of harm. 

From our perspective, the pursuit of speed and scale in war that are behind these developments raise critical humanitarian, ethical, and legal concerns. It is unacceptable and dehumanising if decisions to use force are made based on no real deliberation or judgment on the outputs of automated systems. People and their judgment are not perfect, but taking away moral engagement and deliberation risks atrocity, impunity, and indiscriminate harm. The pursuit of speed through technological applications can accelerate the pace of errors and harm, taking warfare beyond the limits of human cognition. We also have many concerns about the data used by AI in the kill chain and how it is processed and systems are trained, including the use of data gathered through mass surveillance, and the encoding and amplification of bias and discrimination such as sexism, racism, and ableism. Users who do not have a good understanding of the data used in systems may not be able to make a meaningful evaluation of its outputs. The principle that systems must always be predictable, reliable, explainable, and traceable enjoys considerable support within the AWS discussion—this should also apply to military AI more broadly, and isn’t possible with current forms of probabilistic AI. And currently, we are already seeing a lack of transparency around how AI systems are being used in warfare and their impacts.

Recommendations 

Until rules, standards and guarantees are put in place to address the grave concerns raised by AI in the use of military force, Stop Killer Robots is now calling for states to put in place:

• An immediate moratorium on using any AI system, including generative AI, as the basis for targeting decisions. We believe some conflict parties have been using AI to inform who lives and dies and what targets should be struck without meaningful deliberation on the outputs produced by systems. This is dehumanising and reckless automation of violence is unacceptable and must stop now;
• We also need immediate transparency from states on the role of AI in the use of military force, and on its impacts. With reports emerging of grave humanitarian consequences, we think it is up to states to demonstrate the safety, legality, and acceptability of their tools and practices, not for us to trust that they will do the right thing.

To prevent humanitarian consequences, ensure accountability, and meet the range of legal and ethical challenges raised by AI and automation in military systems, the international community must then take all necessary steps to: 

• Prevent, in systems that contribute to the use of force, the use of data gathered in ways that infringe on human rights and other bodies of international law, as well as the use of unreliable or otherwise inappropriate data; 
• Ensure that all systems that are used in military operations are operated with meaningful human control—which includes ensuring that tasks are within the cognitive capacities of users; 
• Prevent the de facto automation of life-and-death decisions, and ensure that there is always deliberative and informed decision-making in the use of force; 
• Require predictability, reliability, explainability, and traceability in all systems, as well as transparency about their uses and impacts. 

There is widespread concern amongst civil society about the hugely harmful impacts on people that developments in military AI are already having, and agreement about the responsibility of states and companies to respond to and prevent these. A broad range of academic and technical experts, as well as civil society from around the world, launched a Joint Statement on AI in Warfare on the first day of the UN talks on AI in the military domain. Stop Killer Robots endorsed this statement, which elaborates a range of concerns and calls for a halt to the use of AI tools in military targeting, adherence to international law, and transparency. 

Humanitarian and ethical concerns are often not strongly reflected in the more technical discussions that take place in meetings such as these. However, they should be at the centre of states’ efforts to formulate normative and regulatory responses. We strongly encourage the international community to put the risks of harm to people, our dignity and autonomy, and our environment at the forefront of both their discussions on military AI and broader efforts to address the dangers posed by the rise of AI and autonomy in the use of force.

It’s clear that a much stronger international response is needed, and that we need to start making rules now.

A version of this article was first published in Reaching Critical Will, AWS Diplomacy Report, Vol. 3, No. 1, Available at: https://reachingcriticalwill.org/disarmament-fora/others/informal-exchange-militaryai/aws-dr/18210-aws-diplomacy-report-vol-3-no-1