Human control is the essential common goal that allows allies to develop AI standards and norms.
Stanley-Lockman, 2021 - Center for Security and Emerging Technology [Zoe CSET Issue Brief August “Responsible and Ethical Military AI Allies and Allied Perspectives” https://cset.georgetown.edu/publication/responsible-and-ethical-military-ai/ Acc. 4/21/22 TA]
In this light, this issue brief seeks to provide policymakers and analysts with one view on how similarities between allied perspectives on ethical AI for defense create opportunity for increased collaboration, and how the differences that are beginning to take shape can undermine said collaboration. Alignment and collaboration start with an understanding of variations in definitions of terms like trustworthy AI, ethical AI, and responsible AI in the defense context. These definitions are often fluid, depending on the legal, ethical, and cultural traditions of different countries. But different conceptions of responsible military AI nevertheless share foundations that help frame the analysis here. Broadly speaking, this issue brief focuses on how defense stakeholders steward AI innovation and integration in ways that: (1) respect the moral and ethical reasoning that underpins the responsible use of force; (2) meet and enhance compliance with law, which is based on ethics and translates reasoning into concrete obligations; and (3) minimize risks and unintended consequences for a safer and more secure international security environment. To uphold ethical, legal, and safe foundations of AI development and deployment in defense, allies coalesce around two shared themes in their approaches to responsible military AI. First is that decisions around the design, development, deployment, and diffusion of AI do not enter into a vacuum, but rather into an existing, multi-layered legal framework. It is not controversial for democratic countries to declare their shared obligation to respect law in order to remain accountable.8 This accountability is owed to domestic citizenries, to the armed forces themselves, and to allies and partners in coalition settings, as well as to adversaries and the international community at large. As such, for some allies, emerging conceptions of responsible AI are closely interlinked with responsible state behavior, with continued legal compliance as the minimum requirement.9 The second commonality in all of the frameworks examined here is a shared focus on human centricity. There are several definitions of human-centric AI, but for the purpose of this analysis, it can be understood as the idea that AI is designed to meet human needs and improve upon the role of the human.10 Not all frameworks use the term itself, but all stress the central role of humans in that machines should not replace humans and that humans remain responsible and accountable for decisions. By extension, this means a common approach to designing AI systems in such a way that the human user is not expected to adjust her or his own decision-making capacities to conform to the technology.11 The inverse would place the machine at the center of decision-making systems. Meanwhile, countries consider humans to be central to defense planning and operations, and stress in their positions that they do not think it moral or lawful to delegate human responsibility to machines.