presentsUNAIVERSE
The UN AI timeline

    Learn AI /AI Governance and UN Terminology /Lethal autonomous weapons systems

    Lethal autonomous weapons systemsLAWS

    Also called: LAWS, autonomous weapons, killer robots.

    Detail

    In five seconds

    What it is

    Weapons that can select and apply force to targets with reduced or no direct human control at the moment of attack, raising legal, ethical and security concerns.

    In plain words

    These are weapons where software and sensors may take on part of the job of finding a target and using force, sometimes with less direct human control in the moment.

    An everyday example

    Two offices use the same email filter. In the first, it flags messages that appear to be duplicates, but a colleague decides whether to delete each one. In the second, the filter deletes anything matching its rules as soon as it arrives. Someone set the rules earlier, but no one makes the final decision for each message. This illustrates the difference between advance configuration and human control at the moment of selection and action. Deleting email is, of course, nothing like applying force, but the contrast helps show why the timing of human judgement matters.

    Where the story stops working

    A grain-store gate is a civilian safety device, not a weapon, and the harm it causes is far lower than armed force. Real weapons operate in much more complex environments, can kill people, and are governed by international law and military command structures. The analogy helps show the difference between preset rules and real-time human control, but it does not capture battlefield uncertainty, legal review, or the scale of harm.

    Remember thisThe issue is not software alone. It is whether, after activation, a weapon can select and apply force to targets with limited or no direct human control, under a definition that remains disputed.

    This term is contested

    Yes. The scope of LAWS is genuinely disputed across states, experts and advocacy groups, especially around how much autonomy, what functions, and what level of human control place a system inside the term.

    Understand it

    Understand it

    Why it matters

    This term matters because it affects who controls life-and-death force, how accountability works, and what rules states may negotiate. A non-specialist does not need to master weapons engineering to ask the right questions: Who selects the target? Can a human intervene in time? How predictable is the system in real conditions? If the answer is unclear, legal review, human rights concerns, civilian protection and strategic stability all become harder. The term also matters because public debate often mixes precise policy language with advocacy language, which can create confusion about what exactly is being regulated or prohibited.

    How automation differs from a more autonomous attack chain

    After viewing, the reader should be able to say why not all automated weapons are discussed as LAWS, and where human control can be reduced.

    1. 01

      Shared starting point

      Both columns begin with sensors gathering information from the environment, such as images, radar or heat signals.

      Place identical sensor icons at the top of both columns feeding into the next box with solid arrows.

    2. 02

      Column A: automation

      A human operator reviews the information, decides what the target is, and authorises force. The system then performs limited tasks such as tracking or stabilising aim.

      Left column: boxes labelled Sensors → Human identifies target → Human authorises strike → System executes preset function. Highlight the two human boxes in one colour.

    3. 03

      Column B: higher autonomy in critical functions

      Software processes the sensor data, classifies a possible target and may trigger force under preset rules, with reduced or no direct human approval at that moment.

      Right column: boxes labelled Sensors → Software classifies object/behaviour → System selects target under rules → Force applied. Show a smaller side box labelled Human sets parameters earlier connected with a dashed line.

    4. 04

      Why the distinction matters

      The policy debate centres on the critical functions of selecting and attacking targets, not on every kind of military automation.

      Add a bracket around the right-column middle boxes labelled Critical functions debated in LAWS discussions. Add a note below left column saying Not every automated system falls here.

    A two-column comparison. Left column shows an automated weapon with a human choosing the target and the system only carrying out narrow preset actions. Right column shows a more autonomous chain where sensors, software and attack functions are linked with less direct human action at the moment of strike. Read top to bottom in each column, then compare the human role.

    Hover or tap each box to reveal a short note: sensor input, target selection, human approval, preset execution, or force application.

    Describe this without the diagram

    This diagram compares two pathways. In both, sensors first collect information. In the left pathway, a human operator identifies the target and authorises the strike, and the machine then performs limited preset actions such as tracking. In the right pathway, software analyses sensor data, classifies a possible target, and the system selects and attacks under preset rules, while a human may only have set parameters beforehand. The lesson is that the LAWS debate focuses on reduced human control over selecting and attacking targets, not on all automation.

    With reduced motion: A static two-column chart with the same boxes and labels, no animation. Colour and text alone show where the human acts directly in the left column and indirectly or earlier in the right column.

    This term covers several related labels, not one perfectly settled category. “Lethal autonomous weapons systems” or “LAWS” is a formal policy phrase often used in United Nations and arms-control discussions. “Autonomous weapons” is broader and may include systems that use force with varying degrees of machine control. “Killer robots” is a campaign and media phrase, not a precise technical or legal term.

    The basic idea is about a weapon system that receives information from sensors, such as cameras, radar or other signals, processes that information with software, and, after activation, selects and engages a target with limited or no further human intervention. What goes in: sensor data and mission rules. What happens: the system classifies objects or behaviours and links that classification to an attack function. What comes out: a selected target linked to actual force.

    The main point of debate is not whether all military automation counts. Many weapons already automate narrow tasks, such as stabilising aim or intercepting incoming rockets. LAWS usually refers to something more specific: systems with significant autonomy in the critical functions of selecting and attacking targets. That is why this term is often discussed alongside, but not as the same thing as, meaningful human control.

    People disagree about exactly where the line should be drawn. Some focus on whether a human approves each strike. Others focus on the type of target, the operating environment, predictability, or compliance with international humanitarian law. So when you hear the term, it is important to ask: autonomous in what function, under what limits, with what human role, and at what stage of using force?

    How “The shared inbox rule” maps onto the real thing

    In the storyIn the AI system
    Cameras and motion sensors at the gateSensors collecting data about the environment
    Pattern rules for who counts as a threatSoftware logic or models used to classify and select targets
    Shutting the gate and triggering the sprayApplying force or initiating an attack function
    Guard sets rules in advance but may not approve each incidentHuman involvement before deployment rather than direct control at the moment of action

    A worked example

    During a technical briefing, delegates compare two notional configurations of the same defensive system to examine how the human role changes after activation.

    Starting from: Both configurations use sensor data and software to classify objects. In one, an operator must approve each engagement. In the other, the system may select and engage matching objects without further approval.

    1. In the first configuration, the software identifies possible targets and presents them to an operator.
    2. The operator reviews the information, decides whether a target should be selected and separately authorises any use of force.
    3. In the second configuration, the operator activates the system and sets operating constraints, but the software subsequently selects and engages targets without a fresh human decision.
    4. The comparison focuses on who performs the critical functions of target selection and attack. It does not assume that every delegation would classify the second configuration in the same way.

    Result: The hardware may be unchanged, yet the degree and timing of human control can place one configuration more directly within debates on lethal autonomous weapons systems. The applicable definition remains contested.

    In a United Nations workplace

    A disarmament desk officer in a Panama City liaison office is clearing talking points for the chief of section before afternoon informals. Delegations are using several terms for weapon systems that can select and engage targets after activation, and the text must reflect those differences without implying that a definition has been agreed.

    The officer needs to distinguish autonomy in the critical functions of targeting and attack from automation elsewhere in a system. Precise wording matters because positions may turn on the human role, predictability, operating conditions and when human judgement is exercised.

    Before you try thisDo not put confidential negotiation notes, unpublished delegation positions or security-sensitive technical material into an external AI tool unless it is approved for that material. A generated summary must also be checked against the record, particularly where terminology remains disputed.

    Where you may hear it

    • Informals on a draft text concerning autonomous weapons
    • A First Committee side event on human control over force
    • A technical briefing for delegations on military applications of AI
    • The margins of a disarmament meeting in Geneva

    Avoid confusion

    Avoid confusion

    Often confused with

    Common misconceptions

    • Any weapon with software or automatic features counts as a lethal autonomous weapon system.

      Not necessarily. Many systems automate narrow tasks such as navigation, tracking or defence against incoming fire, while a human still selects and authorises the target. LAWS usually refers to autonomy in the critical functions of selecting and attacking targets, though states disagree on the exact threshold.

    • LAWS means fully independent machines that fight without any human involvement at all.

      Some public discussion imagines that extreme case, but many real policy debates concern partial autonomy. A human may choose the mission area or set rules in advance, yet not review each target at the moment force is used. That reduced real-time control is a central concern.

    • ‘Killer robots’ is just another exact technical name for the same thing.

      It is a campaigning and media label, not a precise legal or technical definition. It can be useful in public advocacy, but in negotiations people often prefer narrower terms such as LAWS or autonomous weapons because the scope of regulation may depend on wording.

    Quick check

    Which statement best captures why LAWS is debated separately from ordinary automation?

    Go deeper

    Go deeper

    The simple version

    These are weapons where software and sensors may take on part of the job of finding a target and using force, sometimes with less direct human control in the moment.

    The technical version

    In policy discussions, LAWS generally refers to weapon systems that, once activated, can use sensor inputs and software processing to identify, classify, select and possibly engage targets with limited or no further human intervention in the critical functions of targeting and attack. The exact definition remains disputed. Some formulations focus on autonomy after activation; others focus on the human role in target selection, the predictability of system behaviour, the operational environment, or compliance with international humanitarian law. The term should be distinguished from general automation, remote control, and decision-support systems that do not themselves select and attack targets.

    Reviewed external resources

    Where this came from

    Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.

    The one thing to remember

    The issue is not software alone. It is whether, after activation, a weapon can select and apply force to targets with limited or no direct human control, under a definition that remains disputed.