• Journal Article

Artificial Intelligence in Space: An Analysis of Responsible AI Principles for the Space Domain

T. Graham; K. Thangavel.


Advances in Artificial Intelligence (AI) technologies are enabling a plethora of new applications across many industries. There are already a multitude of applications for AI systems in the space industry, but as the AI and space industries continue to grow in size and value rapidly, further use-cases will become apparent and proliferate to all corners of space operations and data analytics. Such space-based AI systems will bring many economic, scientific, and environmental benefits; however, they could also enable harm to individuals, organisations, and the environment if they are not developed and managed properly. Potential breaches of privacy through AI-assisted analysis of Earth observation imagery and collisions between objects in orbit due to malfunctioning automated maneuvering systems are examples of the harms that could eventuate if poorly designed AI systems are deployed in the space-sector. Responsible AI practices are needed to ensure such risks do not eventuate. ‘Responsible’ (or ‘ethical’) AI has emerged as a discipline designed to guide responsible AI development wherein the goal is to maximise the benefits of AI systems for individuals and society while mitigating against any potential harm that they may cause. Commonly accepted Responsible AI principles include accountability, contestability, fairness, security, privacy, transparency, explainability, and reliability. At times notions of ‘do-no-harm’ and generating ‘net benefits’ for society and the environment are also included. These principles of Responsible AI are generalizable and industry agnostic; however, they should be carefully considered in the context of the unique physical, economic, political, and technological characteristics of the space domain before being adopted wholesale by the space industry. While concepts such as security and reliability can be readily applied to applications of AI systems in the space domain, other ideals such as contestability, fairness, and explainability may not be as relevant to the use cases found within the space industry. This paper introduces the concept of Responsible AI and Responsible AI principles and then examines the applicability and appropriateness of widely accepted Responsible AI principles in the context of existing and emerging regulatory instruments relevant to the space industry. This serves as a first step towards creating a standardized regulatory framework for the responsible development of space-based AI systems and preventing harms associated with such systems occurring.

Read full Publication