AI is reshaping battlefield operations and thereby moving up on governance agendas by the day. The autonomous capabilities of weapon systems in particular have drawn growing attention from the international community for over a decade. Disarmament Week offers a timely opportunity to examine who decides how military AI behaves, and why that matters for reducing risks and reinforcing accountability.
In September 2026, the Group of Governmental Experts on lethal autonomous weapon systems in Geneva agreed on a set of elements for a future instrument on these weapons. Yet, AI-enabled capabilities can take much more diverse forms in and beyond the military domain.
As both a force multiplier and a disruptor on many fronts, AI can enhance or disrupt logistics; sharpen or distort intelligence, surveillance and reconnaissance; strengthen cyber defences or deepen vulnerabilities; and help detect or spread disinformation.
Where industry makes decisive decisions
Long before an AI-enabled system reaches the battlefield, someone decides what data it learns from, how it is tested and how much it is allowed to decide on its own.
That someone is usually not a State but an engineer at a company.
These design choices shape whether a system is reliable, predictable and lawful, and they are made years before any commander uses it. The decisive role of industry extends to the entire system life cycle of most AI-enabled military capabilities, spanning from the design and testing phase to post-deployment maintenance and decommissioning.
Across all stages, technology developers, integrators and providers make crucial decisions on how to source and handle data, how to simulate operational conditions and where to set the degree of autonomy through decision thresholds.
This centrality, and often even primacy, in consequential technical decisions makes the private sector’s role and responsibility a focal point of international governance discussions on military AI. The pace of development also sits primarily in the hands of companies.
As drivers of innovation, they tend to move faster than States can regulate them. That speed puts them closest to new capabilities and therefore first in line to explore, understand and draw leverage from them.
What’s more, the military AI industry itself is quickly evolving alongside the technologies it develops. As demand for their systems grows, new firms are entering the market while established defence companies form new partnerships, acquire AI start-ups and expand AI implementation into their products. Importantly, newer entrants and established defence companies may differ in their institutional experience and in the internal structures they have built for compliance and accountability.
Beyond the narrower defence industry landscape, AI now spans nearly every sector and security domain. While specialized companies focus on dedicated battlefield capabilities, the ecosystem of actors with a stake in the life cycle and supply chains of these systems is growing more complex and interconnected. Involvement of industry players can range from direct procurement, through subcontracting networks, to third-party testing.
Making military AI lawful by design
Translating governance principles for military AI into operational and technical practice is a necessary step towards making these systems “lawful by design“. This means integrating legal requirements as core design parameters from the earliest stages of development, and sustaining compliance across the entire life cycle of these technologies. This approach cannot, however, displace the responsibility of States.
In principle and practice:
- Companies, for their part, have a responsibility to respect human rights that exists over and above legal compliance, as set out in the UN Guiding Principles on Business and Human Rights. They can integrate legal assessments into their risk management frameworks and factor international humanitarian law considerations into design and testing processes.
- States remain bound by international humanitarian law and international human rights law and, as customers, require assurance that the capabilities they acquire enable them to meet these obligations. They can embed responsible AI principles into procurement and export control processes, articulate clear expectations for suppliers, and invest in the technological, human, operational, financial and procedural capacity to apply them. This includes legal reviews of new weapons, means and methods of warfare under Article 36 of Additional Protocol I, which may need to be conducted iteratively given the evolving nature of AI systems.
Multi-stakeholder consultations convened by UNIDIR point to a shared recognition, among States and industry alike, that legal and ethical requirements need to be translated into technical requirements, and that a clear distribution of roles and responsibilities is essential for effective public–private partnerships.
Building a shared baseline with industry
Before companies and governments can act on these measures, there needs to be a shared understanding and expectation of what responsible industry behaviour means. A framework that takes stock of multi-stakeholder perspectives through close consultations on responsibility and ethics, and turns them into clear guidance on accountability and compliance. This would give all involved stakeholders the shared baseline they need.
Industry actors must be included in such a process. UN Secretary-General António Guterres regularly emphasizes the involvement of a range of stakeholders, including industry, in policy initiatives on responsible military applications of AI. Consultation with experts and academia, civil society and other stakeholders, such as the humanitarian sector, is equally crucial. It ensures that popular and diverse expectations are included and galvanized, and that these efforts are credible.
The inclusion of industry perspectives is also vital for ensuring that any frameworks for responsible behaviour remain appropriate, effective and grounded:
- Practical experience. When it comes to translating requirements into system design and engineering choices, industry has the most direct and practical experience of what is possible and where technical constraints are.
- Realistic outcomes. Industry involvement and input ensure that outcomes for AI are realistic, feasible and actionable, and meet the expectations.
- Technical expertise. The private sector can bring valuable technical expertise and understanding States lack to the development process of governance frameworks.
- Buy-in. Ultimately, close consultation with companies increases buy-in for any solutions proposed.
Gaining insight into and maintaining access to the technological frontier is increasingly a priority and integral aspect of national AI strategies of States, spanning from military systems to frontier foundational models.
Why States have a stake in responsible industry behaviour
Frameworks for responsible industry behaviour are thus equally in the interest of States. Swiss Federal Councillor Albert Rösti, who heads the Federal Department of the Environment, Transport, Energy and Communications, has called for companies to set their own guidelines and standards for trustworthy AI, alongside voluntary commitments, even before international agreements or other forms of global regulation materialize.
States are also invested in what these commitments and measures would entail in substance. Military procurement can be a direct mechanism by which States implement legal obligations.
Although many of the critical design choices rest in the hands of companies, international legal responsibility falls on the State that uses a system, not on whoever developed it.
For example, the interpretability of AI outputs is often limited by their “black box” nature of these systems. While warfare will inherently require navigating uncertainties, obscure system logs may add layers of complexity to States’ obligation to conduct effective investigations of alleged violations of international humanitarian law.
In practice, governments need to be clear about the questions they expect industry partners to answer as the basis for partnerships. Explicit expectations and a clear allocation of roles and understanding of compliance responsibility within and beyond contracts, fosters sustainable public–private partnerships and procurement relationships. These give States access to the information and documentation they need, and greater confidence about technical reliability.
UNIDIR’s role in turning expectations into guidance
This is where UNIDIR’s work matters. With its subject matter expertise and convening power, the Institute can help industry, States and other stakeholders unpack these complex questions on responsibility, accountability and compliance in military AI.
Through its Framework of Responsible Industry Behaviour for AI in the Military Domain, UNIDIR’s Centre of Excellence on AI, Peace and Security aims to support industry and States in defining their roles for responsible military AI. Developed with the Office of the UN High Commissioner for Human Rights and in consultation with industry, States, civil society and the humanitarian sector, the framework will translate these expectations into voluntary, practical guidance on what responsible behaviour looks like at each stage of a system’s life cycle.
The work reflects UNIDIR’s commitment to designing inclusive processes that foster greater trust, sustained dialogue and principled technological development in a field crucial to global peace and security discussions.
