2. Rules of Engagement - NATO is Key because it establishes Rules of Engagement which are essential for ethical AI use.
Vestner, 2021 - Head of Security and Law Programme at Geneva Centre for Security Policy [Tobias, July 8 “Warfare and Artificial Intelligence” in Robin Geiß and Henning Lahmann (eds), Research Handbook on Warfare and Artificial Intelligence -forthcoming https://www.gcsp.ch/publications/military-operations-and-artificial-intelligenceGCSP Acc 5/27/22 TA]
ROE are appropriate tools to determine how to use AI under which conditions for specific contexts and missions. ROE – or related rules of behaviour – may set the parameters for diverse military applications of AI, thereby translating given political, military, legal, and ethical considerations and limitations of documents at a higher echelon, such as doctrine or international legal obligations, into concrete instructions. This can represent a framework for action to be programmed into the AI system. For example, ROE could determine a geographical zone or a certain list of potential tasks for which systems are authorized to take action. Outside those limits, they would not act on the processed information. Time checks or other limits, such as pre-set permission to (not) engage specific targets, may also be fixed by ROE.100 Similarly, ROE can foresee that a system needs to flag unexpected events or issues. In this context, some have suggested that AI may be able to choose which ROE to apply based on the environment or its programmed mission.101 ROE can also define the interaction between humans and AI systems for specific missions. In particular, ROE can establish how a commander or operator needs to monitor and control the system during deployment. As the need for human control may vary according to the specific task attributed to an AI system and the respective context and operation, ROE for AI can define the level of autonomy for certain types of operations or phases thereof.102 ROE can further address or refer to other sources, such as manuals and directives, on how to implement various forms of human control, such as direct, share, or supervisory control.103 Importantly, ROE may limit commanders’ or operators’ authority, which may force them to refer up in the chain of command. This can be a significant role of ROE for the human-machine teaming in military operations, notably when confronted with unanticipated situations or issues for which the system or its use had not been previously authorized. ROE are particularly relevant when AI is used for or in relation to targeting as this implies harming persons and objects. Notably if considered that AI cannot incorporate ethical or contextual assessments into its decision process,104 human control and judgment should be meaningful in the context of decisions regarding the use of lethal force.105 While most publicly available policies establish this principle, as described above, they do not specify its precise meaning. ROE and directives can fill this gap. To this end, a code of conduct for operators of AI systems related to targeting or a model of ROE for such systems could be established.106