Safety, Legal and Ethical Implications of Artificial Intelligence in the Military Domain

An Australian Army Perspective

Edited by: Assoc Prof Zena Assaad, Dr Jessica Dorsey, Dr Neil Renic

The integration of artificial intelligence into a range of military systems is no longer the subject of theoretical speculation. States are, at this moment, designing, testing, procuring and deploying AI-enabled and autonomous capabilities across all battlefield domains. These technological and military advances are taking place within an insecure geopolitical climate—intensified great power contestation; a resurgence of civilisational politics; destabilising arms racing; and the return of large-scale, inter-state war. Amidst such conditions, calls are growing for middle powers like Australia to evaluate their approach to military AI—accelerate innovation, compress regulatory processes, and optimise warfighting to better position itself against the security challenges facing the region.

This narrative leaves a lot unaddressed, including a range of questions with genuine moral weight. Firstly, is the military AI arms race as critical and seemingly inevitable as frequently described? If it is, then what does this actually license Australian policymakers to do in this space? Are existing technical safeguards sufficient or in need of reform? What are the unintended but foreseeable negative effects of speeding and scaling up our technological acquisition? Within the increasingly systematic, data-infused, sensor-saturated battlefield we are building towards, can we realistically maintain human operators as something more than morally inert actioning devices? These questions do not fall away, regardless of how seemingly dire the political and military stakes may be.

Just as importantly, who ought to bear the costs for the choices we are making; for our greater appetite for innovation, experimentation and risk; for the uncertainties we are accepting, and in some cases, magnifying? Too often in war, it is the innocent. Responsible governance of this technology requires meaningful reflection about the innocent civilians most likely and least able to bear the costs when our military-technological policies go awry.

There are also considerations around how existing legal frameworks can be applied to technologies that are changing the circumstances in which military decisions are made. How legal requirements are to be interpreted and implemented as decision-making becomes increasingly dependent on systems that can process vast quantities of data, generate recommendations, and in some circumstances act with limited (or no) human intervention is becoming more critical by the day. These developments call into question the meaning and practical operation of human judgement and control, and make us critically inquire how legal responsibility can be maintained when the processes through which decisions are reached become increasingly complex, distributed, and technologically mediated.

For Australia, these problems are further shaped by the increasingly collaborative nature of defence technology development, including through AUKUS, a factor addressed in this special issue of the Australian Army Journal. The development and integration of AI-enabled capabilities with allies and partners presents significant opportunities, while also creating challenges concerning interoperability, differing legal and regulatory requirements, assurance standards, and the allocation of responsibility across national and organisational boundaries. More fundamentally, AI integration may alter the ‘choice architecture’ within which military decisions are made, in so doing shaping which options are identified, how information is presented, the speed at which decisions must be taken, and the opportunities available for appropriate levels of human intervention and compliance with international law. The legal implications therefore extend beyond the question of whether particular uses of AI are lawful. These systems also introduce socio-technical implications concerning whether the systems, processes and institutional arrangements through which military decisions are made are capable of sustaining the exercise of lawful judgement and preserving meaningful accountability at various levels as the character of military decision-making changes.

The contributions in this special issue take seriously these questions, detailing how the Australian military can build meaningful assurance into AI-enabled systems across the entire lifecycle: research and design, procurement, testing, legal review, deployment, and use. These contributions are not agnostic as to the hard choices and difficult trade-offs facing Australia within this security environment. But nor do they accept that the same pressures justify an abandonment of safety, legal, moral and political considerations.