CBMs can reduce the risk of AI instability through transparency and monitoring.
Horowitz and Scharre 2021 - Director of the Emerging Capabilities Policy Office in the Office of the Under Secretary of Defense for Policy and Vice President and Director of Studies at CNAS [Michael, Paul, January 12 2021, “AI and International Stability: Risks and Confidence- Building Measures”, https://www.cnas.org/publications/reports/ai-and-international-stability-risks-and-confidence-building-measures, Acc 6/18/22. M.A.]
The Role of Confidence-Building Measures AI potentially generates risks for international security due to ways AI could change the character of warfare, the limitations of AI technology today, and the use of AI for specific military missions such as nuclear operations. Especially given the uncertain technological trajectory of advances in AI, what are options to reduce the risks that military applications of AI can pose to international stability? To advance the conversation about ensuring that military AI adoption happens in the safest and most responsible way possible, this paper outlines a series of potential confidence-building measures aimed at mitigating risks from military uses of AI. We introduce these ideas as preliminary concepts for future research, discussion, and examination, rather than to specifically advocate for any of these options. But progress in mitigating the risks from military AI competition requires moving beyond the recognition that risk mitigation is important to the hard work of suggesting, evaluating, and examining the benefits and drawbacks of specific mechanisms. This paper focuses on confidence-building measures, a broad category of actions that states can take to reduce instability risks. CBMs include actions such as transparency, notification, and monitoring designed to reduce various risks arising from military competition between states. They generally encompass four areas, as Marie-France Desjardins describes: Information-sharing and communication Measures to allow for inspections and observers “Rules of the road” to govern military operations Limits on military readiness and operations Confidence-building measures are related to, but distinct from, arms control agreements. Arms control encompasses agreements states make to forgo researching, developing, producing, fielding, or employing certain weapons, features of weapons, or applications of weapons. The set of possible actions states could take is broad, and this paper will focus on the potential benefits and drawbacks of specific AI-related confidence-building measures. Arms control for military AI applications is a valuable topic worthy of exploration, but beyond the scope of this paper.