23 March 2023
IP23030 | Inaugural Summit on Responsible AI in the Military Domain: Limitations and Proposed Pathways
The first global summit on Responsible AI in the Military Domain was held in The Hague in February 2023 to discuss how the military can employ AI responsibly. It culminated in a joint call to action, which addressed the urgent need for a military AI governance framework. While the document demonstrates a significant international effort to find ways to use military AI responsibly, it faces several limitations. WICHUTA TEERATANABODEE suggests several steps to maximise the potential of the next summit in finding solutions for the responsible use of AI in the military domain.
COMMENTARY
Artificial intelligence (AI) uses algorithms to mimic human intelligence and is intended to enhance the speed, precision, and effectiveness of human efforts. AI-enabled technologies have been employed in various sectors, including the military.
While AI offers the advantage of helping to minimise human error and increasing the effectiveness of military operations, it comes with potential risks and harms. Due to concerns over the urgent need for norms and governance frameworks for the military use of AI, in February 2023, the Netherlands and South Korea jointly convened the Responsible AI in the Military Domain (REAIM) Summit, which concluded with the “REAIM 2023 Call to Action”.
The REAIM 2023 Call to Action
The Call to Action was drafted and endorsed by over 50 states, including those boasting the world’s leading AI technologies, such as the United States, China, France, Germany and Japan, as well as Singapore.
The preamble of the document acknowledges the potential of AI, yet also recognises that the rapid adoption of AI in the military domain brings about visible and invisible challenges. On the one hand, the failure to deploy AI in a timely manner, especially in this time of fierce strategic competition, may result in a military disadvantage. On the other hand, a premature adoption of AI without sufficiently well-informed research, testing, and assurance may lead to unintended and harmful outcomes.
Furthermore, the document acknowledges that, despite continuous efforts to learn about AI, we do not and cannot fully comprehend and anticipate the implications and challenges resulting from AI applications within the military domain. This is partly due to the distributed nature of military decision-making and the diverse nature of the AI ecosystem, where different sectors are involved throughout the entire life cycle of AI – from design to development and deployment.
The document accordingly puts forward several points calling to action. It stresses the significance of a holistic, inclusive, and comprehensive approach in addressing the possible impacts, opportunities, and challenges of the use of AI in the military. This effectively means there is a need for collaboration and information exchange between relevant stakeholders, including governments, the private sector, civil society, and academia.
Along with stressing the need for an international effort to govern AI, the document invites all states to increase general comprehension of military AI through research, training courses and capacity building activities, as well as to develop national frameworks, strategies, and principles for responsible AI in the military domain.
More Needs to be Done
The REAIM 2023 Call to Action demonstrates a significant effort to develop further collaboration and reach potential agreement on shared practices for military AI, an attempt worthy of recognition. However, the document has three key limitations: it lacks concrete plans; does not consider the diversity of military AI capabilities; and did not have a multi-stakeholder document-drafting process.
First, it is understandable that finding agreements, even non-binding ones, is a difficult task at any international forum, especially when given a limited timeframe. However, merely inviting or welcoming states to cooperate or develop national strategies on responsible AI in the military domain might not be strong enough, especially considering the urgency of the issue. Furthermore, as preparations for the second REAIM Summit are already under way in South Korea, participants, particularly state representatives, should have taken advantage of this continuation to set more concrete and trackable call-to-action proposals.
For example, the proposal could have included urging country representatives who participated in or supported the drafting of the Call to Action to come up with a roadmap or plan towards responsible AI in the military domain for their respective countries and/or regions. The second summit could then serve as a platform to follow up with each state’s progress on the development of national or joint strategies.
Second, the content of the Call to Action is broad and ambiguous. There is a general acceptance that AI has been applied in a wide range of military fields, from transport and logistic systems to decision support systems, autonomous systems, and even killer robots. The capabilities of each of these applications vary. Depending on the means and purposes of military AI applications, the ends come with different levels of contentiousness and impacts for civilians. Thus, concrete frameworks to govern killer robots, for example, are likely to be, and should be, different from those created for military logistic support systems.
Third, while the importance of a multi-stakeholder approach, characterised by the involvement of relevant stakeholders, was frequently underlined, both at the summit and specifically in the Call to Action, these actors did not take part in discussing or drafting the document. This constituted a missed opportunity to create the holistic, inclusive, and comprehensive governance framework stressed in the Call to Action.
The private sector overseeing the developing and testing processes of AI, for instance, would have been able to offer valuable opinions on the feasibility of potential plans towards making military AI more responsible. Likewise, international and non-governmental organisations might have been able to offer ideas from civil society and human rights perspectives that might have been missed during the process.
Paths Forward
Based on the above reflections on the Call to Action, this paper recommends that the next summit consider the following three steps, particularly involving the process of discussing and drafting a new call to action.
First, participants, particularly the government representatives who are at the forefront of shaping standards and norms on AI, should aim to create more concrete, achievable and trackable plans. These could include, for instance, finding a common understanding of good and responsible practices in the use of AI. Drawing up a clear roadmap for the development of potential AI governance in the military, both at the national and international levels, should also be on the agenda.
Second, the discussions on responsible military AI should account for different ways and potential outcomes of military AI applications. These could include, for example, outlining possible scenarios of military AI applications and determining their level of contentiousness and autonomy, as well as their potential impacts on civilians and the broader society. Spelling out these contexts could help set the foundation for less ambiguous and, consequently, more effective military AI governance.
Third, as REAIM stresses the importance of an inclusive multi-stakeholder approach to responsible AI, it could be beneficial to involve the different stakeholders present at the summit, including the private sector and NGOs, in drafting the next call to action.
A global dialogue such as REAIM is essential for actors to meet and discuss the responsible use of AI in the military domain. This paper is not meant to criticise the REAIM effort but, instead, to help ensure that the next summit can reach its full potential and catch up with the frenetic pace of technological development.
Wichuta TEERATANABODEE is a Senior Analyst in the Military Transformations Programme of the Institute of Defence and Strategic Studies (IDSS), S. Rajaratnam School of International Studies (RSIS).
The first global summit on Responsible AI in the Military Domain was held in The Hague in February 2023 to discuss how the military can employ AI responsibly. It culminated in a joint call to action, which addressed the urgent need for a military AI governance framework. While the document demonstrates a significant international effort to find ways to use military AI responsibly, it faces several limitations. WICHUTA TEERATANABODEE suggests several steps to maximise the potential of the next summit in finding solutions for the responsible use of AI in the military domain.
COMMENTARY
Artificial intelligence (AI) uses algorithms to mimic human intelligence and is intended to enhance the speed, precision, and effectiveness of human efforts. AI-enabled technologies have been employed in various sectors, including the military.
While AI offers the advantage of helping to minimise human error and increasing the effectiveness of military operations, it comes with potential risks and harms. Due to concerns over the urgent need for norms and governance frameworks for the military use of AI, in February 2023, the Netherlands and South Korea jointly convened the Responsible AI in the Military Domain (REAIM) Summit, which concluded with the “REAIM 2023 Call to Action”.
The REAIM 2023 Call to Action
The Call to Action was drafted and endorsed by over 50 states, including those boasting the world’s leading AI technologies, such as the United States, China, France, Germany and Japan, as well as Singapore.
The preamble of the document acknowledges the potential of AI, yet also recognises that the rapid adoption of AI in the military domain brings about visible and invisible challenges. On the one hand, the failure to deploy AI in a timely manner, especially in this time of fierce strategic competition, may result in a military disadvantage. On the other hand, a premature adoption of AI without sufficiently well-informed research, testing, and assurance may lead to unintended and harmful outcomes.
Furthermore, the document acknowledges that, despite continuous efforts to learn about AI, we do not and cannot fully comprehend and anticipate the implications and challenges resulting from AI applications within the military domain. This is partly due to the distributed nature of military decision-making and the diverse nature of the AI ecosystem, where different sectors are involved throughout the entire life cycle of AI – from design to development and deployment.
The document accordingly puts forward several points calling to action. It stresses the significance of a holistic, inclusive, and comprehensive approach in addressing the possible impacts, opportunities, and challenges of the use of AI in the military. This effectively means there is a need for collaboration and information exchange between relevant stakeholders, including governments, the private sector, civil society, and academia.
Along with stressing the need for an international effort to govern AI, the document invites all states to increase general comprehension of military AI through research, training courses and capacity building activities, as well as to develop national frameworks, strategies, and principles for responsible AI in the military domain.
More Needs to be Done
The REAIM 2023 Call to Action demonstrates a significant effort to develop further collaboration and reach potential agreement on shared practices for military AI, an attempt worthy of recognition. However, the document has three key limitations: it lacks concrete plans; does not consider the diversity of military AI capabilities; and did not have a multi-stakeholder document-drafting process.
First, it is understandable that finding agreements, even non-binding ones, is a difficult task at any international forum, especially when given a limited timeframe. However, merely inviting or welcoming states to cooperate or develop national strategies on responsible AI in the military domain might not be strong enough, especially considering the urgency of the issue. Furthermore, as preparations for the second REAIM Summit are already under way in South Korea, participants, particularly state representatives, should have taken advantage of this continuation to set more concrete and trackable call-to-action proposals.
For example, the proposal could have included urging country representatives who participated in or supported the drafting of the Call to Action to come up with a roadmap or plan towards responsible AI in the military domain for their respective countries and/or regions. The second summit could then serve as a platform to follow up with each state’s progress on the development of national or joint strategies.
Second, the content of the Call to Action is broad and ambiguous. There is a general acceptance that AI has been applied in a wide range of military fields, from transport and logistic systems to decision support systems, autonomous systems, and even killer robots. The capabilities of each of these applications vary. Depending on the means and purposes of military AI applications, the ends come with different levels of contentiousness and impacts for civilians. Thus, concrete frameworks to govern killer robots, for example, are likely to be, and should be, different from those created for military logistic support systems.
Third, while the importance of a multi-stakeholder approach, characterised by the involvement of relevant stakeholders, was frequently underlined, both at the summit and specifically in the Call to Action, these actors did not take part in discussing or drafting the document. This constituted a missed opportunity to create the holistic, inclusive, and comprehensive governance framework stressed in the Call to Action.
The private sector overseeing the developing and testing processes of AI, for instance, would have been able to offer valuable opinions on the feasibility of potential plans towards making military AI more responsible. Likewise, international and non-governmental organisations might have been able to offer ideas from civil society and human rights perspectives that might have been missed during the process.
Paths Forward
Based on the above reflections on the Call to Action, this paper recommends that the next summit consider the following three steps, particularly involving the process of discussing and drafting a new call to action.
First, participants, particularly the government representatives who are at the forefront of shaping standards and norms on AI, should aim to create more concrete, achievable and trackable plans. These could include, for instance, finding a common understanding of good and responsible practices in the use of AI. Drawing up a clear roadmap for the development of potential AI governance in the military, both at the national and international levels, should also be on the agenda.
Second, the discussions on responsible military AI should account for different ways and potential outcomes of military AI applications. These could include, for example, outlining possible scenarios of military AI applications and determining their level of contentiousness and autonomy, as well as their potential impacts on civilians and the broader society. Spelling out these contexts could help set the foundation for less ambiguous and, consequently, more effective military AI governance.
Third, as REAIM stresses the importance of an inclusive multi-stakeholder approach to responsible AI, it could be beneficial to involve the different stakeholders present at the summit, including the private sector and NGOs, in drafting the next call to action.
A global dialogue such as REAIM is essential for actors to meet and discuss the responsible use of AI in the military domain. This paper is not meant to criticise the REAIM effort but, instead, to help ensure that the next summit can reach its full potential and catch up with the frenetic pace of technological development.
Wichuta TEERATANABODEE is a Senior Analyst in the Military Transformations Programme of the Institute of Defence and Strategic Studies (IDSS), S. Rajaratnam School of International Studies (RSIS).
