1. This protocol relates to the use of autonomous weapons systems, defined herein, on land, sea and air.
2. This protocol applies only to autonomy in the critical functions of targeting as defined herein and to the functions of activation and deactivation as defined herein.
3. Autonomy in other functions such as navigation and refuelling is not regulated by this protocol.
1. "Autonomy" and "autonomous" refer to systems that are capable of operating in a real-world environment without external human control for an extended period of time.
2. The "critical functions of targeting" are 1) defining targets, 2) selecting targets, 3) engaging targets.
3. "Autonomous Weapons System" (AWS) means a weapons system that has autonomy in one or more of the critical functions of targeting.
4. "Defining targets" means defining what classes of objects the autonomous weapon selects and engages.
5. "Selecting targets" means sensing and confirming objects meets the defined targeting criteria and aiming at them.
6. "Engaging targets" means firing on or using force against the selected targets.
7. "Activation" means turning on the AWS and sending it into offensive combat or enabling its defensive combat function.
8. "Deactivation" means withdrawing the AWS from combat and turning it off.
9. "Responsible Officer(s)" means the human or humans who assume responsibility for the configuration, fitness for purpose and state of repair of the AWS and who can be held accountable for its actions between activation and deactivation.
10. "Fully autonomous" means an AWS that has no humans involved in any of the critical functions of targeting. A fully autonomous AWS defines, selects and engages targets with no external human control.
1. AWS that are fully autonomous as defined in Article 2.10 are prohibited.
2. Autonomy in the function of defining targeting criteria is prohibited. The Responsible Officer(s) must understand what classes of targets the AWS will attack and be satisfied that the AWS can conform to IHL in such attacks.
3. An AWS may not be activated without at least one Responsible Officer assuming responsibility for its targeting configuration, fitness for purpose and state of repair.
4. Autonomy in the activation function is not permitted. A human must activate an AWS after a Responsible Officer has assumed responsibility for its configuration, fitness for purpose and state of repair.
1. Autonomy in the select and engage functions is permitted. However, it is recommended that where practical that there be a human officer who monitors the operation of the AWS and can either confirm or abort targeting decisions made by the AWS in real time.
2. Autonomy in the deactivation function is permitted.