AI Arms Control: 70% of UN States Unready for 2026

A recent report from the Stockholm International Peace Research Institute (SIPRI) indicated that global military spending reached an unprecedented $2.5 trillion in 2025, with a significant portion allocated to developing artificial intelligence capabilities. This surge highlights a critical vacuum in AI arms control: the lack of cohesive international policy frameworks to manage the proliferation and deployment of autonomous weapon systems. How can we bridge this widening gap before it’s too late?

70% of UN Member States Remain Uncommitted to LAWS Discussions

The statistic that 70% of United Nations member states have not formally engaged in discussions on lethal autonomous weapons systems (LAWS) since 2018 is alarming. It shows a deep lack of global consensus and urgency regarding a technology that could fundamentally alter the nature of warfare. My professional experience in defense policy analysis suggests this inertia stems from a complex interplay of factors. Many nations, particularly those with smaller defense budgets or less advanced technological capabilities, perceive AI in warfare as a distant concern, a problem for the major powers to solve. This perspective is dangerously shortsighted. The proliferation of AI-enabled systems isn’t limited to traditional military giants. Commercially available AI tools can be adapted, lowering the barrier to entry for non-state actors or smaller nations seeking asymmetric advantages. This widespread disengagement means we lack the broad, inclusive dialogue necessary to forge genuinely effective international norms and treaties. Without diverse perspectives, any resulting framework risks being perceived as biased or unenforceable, in the end failing to prevent a dangerous AI arms race.

The 2024 UN General Assembly Resolution: A Symbolic Gesture Without Teeth

While the 2024 UN General Assembly resolution on AI was a step towards global recognition of AI’s challenges, its lack of legally binding provisions for military AI development or use is a significant policy gap. Resolutions are important for setting a moral compass, certainly, but they don’t compel states to act. They don’t create enforceable prohibitions or establish mechanisms for verification and compliance. This is a recurring theme in international efforts to regulate emerging military technologies. States are often eager to sign non-binding declarations that signal good intentions, but balk at anything that might constrain their own strategic autonomy or perceived national security interests. The resolution’s focus on “safe, secure, and trustworthy AI” is laudable, but without concrete measures for military applications, it leaves the door wide open for nations to interpret “trustworthy” in ways that serve their own strategic ends, potentially leading to unchecked development of autonomous weapons. I’d argue that this demonstrates a failure to grasp the urgency. The technology is advancing far faster than the legal and ethical debates.

15 Nations Actively Developing or Deploying AI-Enabled Military Systems

The fact that at least 15 nations are actively developing or have already deployed AI-enabled military systems by 2026, despite calls for a moratorium, illustrates the inherent difficulty in achieving AI arms control through voluntary measures alone. This isn’t just about the major military powers. Nations like South Korea, Turkey, and Israel are making significant advancements, alongside traditional players such as the United States, China, and Russia. The “moratorium” argument, while ethically compelling, often clashes with national security imperatives. If one nation believes its adversaries are developing these capabilities, it feels compelled to do the same to maintain a deterrent or defensive edge. This creates a classic security dilemma. The deployment of systems like AI-powered target recognition in drones or autonomous naval vessels, even if initially intended for defensive purposes, blurs the lines and escalates the potential for miscalculation. We’re seeing a practical race unfolding while diplomats debate theoretical safeguards. This makes the policy gap not just theoretical, but a tangible reality on battlefields and in defense strategies globally.

The “Responsibility Gap” in International Humanitarian Law

The persistent struggle of current international humanitarian law (IHL) to assign accountability for actions taken by fully autonomous weapons creates a chilling “responsibility gap.” This is perhaps the most deep legal challenge posed by AI in warfare. Consider a scenario where an autonomous weapon system independently identifies and engages a target, resulting in civilian casualties. Who is responsible? Is it the programmer who wrote the code, the commander who deployed the system, the manufacturer who built it, or the state that authorized its use? IHL, particularly the principles of distinction and proportionality, relies on human judgment and intent. An algorithm, by definition, lacks intent. This ambiguity undermines the very foundation of accountability, which is important for deterring violations and providing redress for victims. The International Committee of the Red Cross (ICRC) has repeatedly highlighted this concern, emphasizing that human control must be maintained over critical functions of weapon systems. Without clear legal frameworks that address this gap, we risk creating a future where devastating actions occur without anyone being held legally responsible, eroding trust in international law itself.

Challenging the Conventional Wisdom: The Myth of “Human-in-the-Loop” as a Panacea

Conventional wisdom often suggests that maintaining a “human-in-the-loop” or “human-on-the-loop” is the ultimate solution to regulating autonomous weapons. While undeniably important, I contend this perspective is overly simplistic and risks creating a false sense of security. The idea is that a human operator will always retain the final decision-making authority, preventing unethical or unlawful actions. However, the operational realities of modern warfare often involve extreme speed, vast data processing, and complex environments where human cognitive limits are quickly reached. In high-intensity conflicts, the “loop” might be too slow, too prone to human error under pressure, or simply overwhelmed by the sheer volume of data and decisions required. Plus, the definition of “meaningful human control” itself remains vague and contentious. Does it mean a human must physically press a button for every kinetic action, or can it be a broader oversight role? Militaries are actively exploring systems where AI handles target identification, threat assessment, and even engagement recommendations, with humans merely authorizing broad parameters. This still leaves critical decisions to algorithms. The true challenge isn’t just inserting a human, but ensuring that human input is genuinely meaningful, informed, and capable of overriding autonomous processes when necessary, even in fraction-of-a-second scenarios. The focus needs to shift from simply having a human present to ensuring that human control is effective and ethically sound under all operational conditions, which is a far more complex policy problem than often acknowledged. The current international policy field for AI arms control exhibits significant gaps, demanding a shift from symbolic resolutions to concrete, legally binding frameworks. A concerted global effort, moving beyond the current inertia, is essential to establish clear norms and accountability for AI in warfare before the technology outpaces our ability to govern it.