Technology Outpaces Diplomacy
The central challenge in regulating military AI is speed. Artificial intelligence is one of the fastest-moving technologies in history, with breakthroughs in machine learning and data processing occurring far quicker than the staid pace of international
treaty negotiations allows. By the time diplomats agree on the wording of a clause, the very technology they are trying to regulate may have already become obsolete or evolved into something new entirely. Discussions at the United Nations, primarily within the Group of Governmental Experts (GGE) under the Convention on Certain Conventional Weapons (CCW), have been ongoing for years. While these talks have produced foundational documents and identified key areas of concern, they have yet to yield a binding international treaty. This creates a dangerous reality where the development and potential deployment of AI-powered weapon systems are advancing in a regulatory vacuum.
The Problem of Definition
Before you can regulate something, you must agree on what it is. For Lethal Autonomous Weapons Systems (LAWS), often called 'killer robots', there is no internationally agreed-upon definition. The debate often centres on the concept of 'meaningful human control'. But what does that truly mean? Does it mean a human must approve every single lethal action? Or is it enough for a commander to set broad 'rules of engagement' for a system that then operates independently? This ambiguity is a significant hurdle. Russia, for example, has argued that without a globally accepted meaning for 'artificial intelligence' itself, any dialogue on regulation is premature. The discussions attempt to use functional characterizations—focusing on what a system can do rather than what it is called—but this still leaves vast grey areas that developers can exploit. The technology ranges from defensive systems that autonomously counter incoming missiles to hypothetical offensive systems that could hunt and engage targets over a wide area without direct human supervision.
A Deep Geopolitical Divide
Even with a clear definition, getting the world's major military powers to agree on restrictions is a monumental task. The United States, Russia, and China—all leading developers of AI technology—have different strategic interests and philosophies. The United States has generally favoured applying existing international humanitarian law, supplemented by national weapons reviews and principles of responsible use, while preserving its ability to develop advanced capabilities. Russia has also resisted a new binding treaty, arguing existing laws are sufficient and using its position in the consensus-based CCW to block language it finds too restrictive. Both nations have been accused of weakening safeguards in recent UN negotiations. Meanwhile, a large coalition of other countries, supported by humanitarian organizations, is pushing for a legally binding instrument to prohibit or strictly regulate LAWS, arguing that existing laws are insufficient for the unique challenges posed by autonomy in warfare.
The Verification and Accountability Nightmare
Effective arms control treaties depend on verification—the ability to ensure all parties are complying. With AI, this is extraordinarily difficult. How can one country inspect another's software code to ensure it complies with rules on 'human control'? The intellectual property involved is often a closely guarded secret of both technology companies and national militaries. Furthermore, AI systems based on machine learning can be unpredictable, their decision-making processes sometimes opaque even to their own creators. This creates a terrifying 'accountability gap'. If an autonomous weapon makes a mistake and violates the laws of war—for example, by attacking a hospital—who is responsible? Is it the commander who deployed it? The programmer who wrote the code? The manufacturer? Without a clear answer, the fundamental principles of accountability that underpin international humanitarian law are threatened.
