In February 2023, the US government released its high-level "Political Declaration on Responsible Military Use of Artificial Intelligence (AI) and Autonomy” at The Hague. This centered on how AI allows machines, including offensive and defensive weapons and weaponry systems, to operate in scenarios that would traditionally require the use of human intelligence. It encompasses facets of AI such as pattern recognition, experiential learning, prediction making, drawing conclusions, and delivering recommendations.
Military AI capabilities are focused on decision support systems that inform the timely decisions made by defense leaders in battlefield situations and strategic planning. They also include the various administrative capacities associated with military operations, from personnel management and accounting, to aspects of intelligence such as surveillance and reconnaissance data capture and processing.
As policy Undersecretary of Defense Sasha Baker described it, the US is a worldwide leader in AI, autonomy, and the responsible military use of technologies. As such, the Department of Defense has been championing ethical AI principles for more than a decade. The political declaration served to advance internal protocol and goals onto the international stage, in recommending norms that extend beyond borders and can be practically implemented by other militaries. While not legally binding, the declaration provides guidelines that enable a responsible military utilization of AI.
Naturally, defining what exactly “ethical AI principles” and “responsible military use” are is a complex and potentially contentious matter. As defined in the declaration, it includes ensuring that AI systems can be objectively audited and that their use is contained within fully defined, legally constrained scenarios. It also involves subjecting such systems to rigorous testing and assessment protocol throughout their lifecycle (with senior-level review reserved for critical, high-consequence applications). Other concrete measures include adhering to international law in AI military use, minimizing unintended bias from creeping in, and using transparent and auditable methods of producing AI systems.
Such AI systems must also have safeguards built in that enable the detection and prevention of behaviors that are outside of the intended scope of use and deployment. This is particularly critical, considering that AI has self-learning and continuous update potentialities built in that, without critical safety features in place, could lead to systems “going rogue” and deploying in unintended ways. Ultimately, this could have consequences that impact relations between nations with heavily armed forces and degrade capabilities to attack and defend.
Deputy Defense Secretary Kathleen Hicks noted that, with such an evaluation, testing, and review system in place, integrating AI in military operations is possible in an expedited and responsible manner. The point of having such technologies in place is enhancing “decision advantage,” as well as maintaining parity with other militaries that possess advanced AI capabilities.
A number of non-legally binding guidelines outlined in the February declaration have been published in the subsequent months, including the November 2nd report “Data, Analytics, and AI Adoption Strategy.” The report recommends moving beyond simply identifying select AI-enabled capabilities for warfighting and instead bolstering the overall organizational environment, through which personnel deploy data analytics and AI capabilities continuously.
In November 2023, the strength of these non-binding principles was bolstered, as the United States government announced that 47 nations, representing military forces around the globe, had signed on to the declaration. This not only places the US at the vanguard of the discussion on how AI can best be implemented by allied and neutral countries, but also allows for best AI practices to be shared and disseminated amongst states that are party to the agreement.
