What this threat is
Lethal autonomous weapon systems (LAWS), sometimes called "killer robots" in public discourse, are systems that can select and engage targets without meaningful human control at the moment of the decision to use force. The key word is "meaningful": some degree of automation has existed in weapons for decades, from heat-seeking missiles to automated naval defense systems. The new concern is AI-powered systems that can make classification and targeting decisions across a wide range of scenarios, with or without a human in the loop.
Modern AI dramatically expands what autonomous weapons can do. Computer vision systems can identify and track individuals or vehicles across large areas. AI can process sensor data faster than any human and make targeting decisions in milliseconds. Drone swarms can coordinate autonomously to overwhelm defenses. These capabilities are being developed and, in some cases, deployed today. Reports from Libya, Azerbaijan, Ukraine, and Gaza have documented weapons systems operating with varying levels of autonomy in identifying and engaging targets.
The concern isn't just about what happens when these systems work as intended. It's also about what happens when they don't. AI systems trained on particular datasets can misidentify targets. They can be deceived by adversarial inputs. They can behave unexpectedly in novel environments. And unlike a human soldier who might hesitate, question an order, or recognize a situation as morally ambiguous, an autonomous system follows its programming. The responsibility for the consequences becomes diffuse and hard to assign.
There's also a strategic dimension. Autonomous weapons could compress decision timescales in conflict to the point where human deliberation becomes impossible. If both sides in a conflict deploy systems that can respond in milliseconds, the risk of accidental escalation, where one autonomous system triggers another's defensive response in a chain reaction neither side intended, becomes real.
Why it matters
International humanitarian law, the body of rules governing armed conflict, requires that any use of lethal force involve discrimination (distinguishing combatants from civilians), proportionality (not causing civilian harm excessive to military advantage), and precaution (taking steps to minimize civilian casualties). These are judgment calls. They require context, discretion, and moral reasoning. Whether an AI system can reliably make those judgments in the chaotic, ambiguous conditions of real conflict is deeply contested among legal scholars and military ethicists.
The proliferation risk is significant. Once a technology exists and proves effective, it spreads. Autonomous weapons that start as tools of major military powers will eventually reach non-state actors, authoritarian regimes, and groups with no interest in following international law. The cost of deploying lethal force drops, and with it the threshold for using it. The potential consequences for civilian populations, for conflict stability, and for the norms that govern warfare are severe.
Where things stand today
Diplomatic efforts to regulate LAWS have been ongoing since 2014 at the UN Convention on Certain Conventional Weapons (CCW). Progress has been slow. Major military powers including the United States, Russia, and China have resisted legally binding prohibitions, arguing that existing international humanitarian law is sufficient or that definitions of "autonomy" are too contested to regulate. A growing coalition of countries and NGOs has pushed for a ban on fully autonomous lethal systems, but no binding treaty yet exists.
The EU AI Act classifies AI systems used in lethal autonomous weapons among its highest-risk applications. Several European nations have adopted national positions calling for a ban. But the states most actively developing these capabilities are not primarily European, and the gap between where governance stands and where technology is moving continues to widen.
How Better Societies helps
Summit: The Better Societies Summit brings together experts on AI governance, international humanitarian law, and conflict to share research and push for coordination. The autonomous weapons problem is a core topic: where the technology is, what governance frameworks are viable, and how to build political will for change.
Compliance: For organizations developing AI systems with potential dual-use or defense applications, EU AI Act compliance is not optional. Our Compliance advisory helps organizations map their systems to the Act's risk tiers and understand what prohibitions and obligations apply.
Accelerator: The Accelerator supports researchers and founders working on AI governance tools, verification technologies, and conflict risk reduction systems. If you're working in this space, we want to connect.