Artificial intelligence is transforming both how wars are fought and how peace can be made. In Ukraine, Sudan, Myanmar, Gaza and Iran, AI is already shaping conflict dynamics across digital and physical frontlines. Beyond active war zones, digital and cyber campaigns target civilians and exacerbate polarization in otherwise peaceful societies.

AI for War and Peace analyzes the ways AI is shaping warfare and geopolitical competition, and how diplomacy, governance and peacebuilding can respond. Part I, AI, Conflict and Competition, covers the AI race between the US and China; AI in the military domain; AI-enabled information wars; AI-enabled cyber conflict; competition over AI infrastructure and resources; and the concentration of power and surveillance capacity that AI enables. Part II, AI and Peace, turns to responses: confidence-building measures and cooperation between the AI superpowers; AI governance at global, regional and national levels; AI in conflict analysis and early warning; AI in diplomacy and peace negotiations; AI for democracy and peacebuilding; and humanitarian AI and ceasefire monitoring.

Outcomes of civil wars differ greatly across countries, even among states with comparable conflict risks — poverty, ethnic cleavages, political instability, and state weakness. Understanding why some civil conflicts terminate through peace agreements or military victories while others fail to end decisively requires attention to the strategic interactions between rebels and states, and how rebel groups mutate and shape competition to reach different conflict outcomes. Yet the critical link between rebel group characteristics — their pre-war organizations, internal dynamics, alliances, and coercive capacities — and conflict outcomes remains under-theorized and empirically underdeveloped.

This book addresses that gap by examining how rebel groups influence civil war outcomes. Drawing on mixed methods — including quantitative analysis of 346 civil conflict outcomes from 1946–2012 and case studies of four conflicts in South and Southeast Asia (Bangladesh, Indonesia, Sri Lanka, and Myanmar) — it demonstrates how rebel features and the number of armed groups shape prospects for victory, defeat, political settlement, or non-decisive outcome. Conflict outcome models incorporating new rebel-group variables predict war outcomes with 83.6 percent accuracy. The study makes two core contributions: first, it integrates non-decisive conflict outcomes into a civil war bargaining framework, demonstrating how rebel organizational changes shape these scenarios; second, it builds new theoretical and empirical links between rebel group lifecycles, the bargaining process, and civil war outcomes. The findings carry strategic implications for states and rebel groups engaged in conflict and for third parties seeking to resolve it.

Dissertation to book project:

Doctoral Dissertation, The Australian National University, May 2024. http://hdl.handle.net/1885/317367

Scores of intrastate conflicts end without military victory or formal peace agreement, yet civil war scholarship lacks a mechanism to explain how these non-decisive outcomes emerge, recur, and endure. Existing "win, lose, or draw" rationalist bargaining frameworks treat settlements as the default outcome when neither party prevails militarily — assuming impasse reliably produces the ripe moment Zartman describes. Yet as conflict systems grow more complex, negotiated settlements continue to decline. This theory note introduces mutually beneficial stalemates — an equilibrium in which conflict parties calculate that maintaining the status quo yields higher returns than renewed fighting or settlements that risk commitment failures — as a mechanism to explain how non-decisive outcomes evolve and are sustained. Where Zartman's mutually hurting stalemate describes a painful impasse generating pressure toward settlement, the mutually beneficial stalemate explains how parties distribute some benefits of peace — limited or no violence, selective cooperation, and hybrid governance arrangements — without the risks of formal resolution. I develop a logic of joint conflict avoidance that shows how parties protect existing gains and preserve organizations against the risks posed by both escalation and resolution. I extend the framework to multi-actor conflicts, where rebel proliferation and external intervention reinforce non-decisive equilibria.

Non-decisive conflict outcomes—instances in which violence ends without a clear victory or formal settlement—are frequent yet understudied scenarios in civil wars. Existing war termination scholarship and rationalist bargaining approaches typically rely on a "win, lose, or draw" framework that fails to capture how conflicts end without clear symbols of termination. This omission limits theoretical understanding of the bargaining processes that produce non-decisive outcomes and narrows the perceived range of rational options available to conflict actors seeking an off-ramp when neither victory nor a negotiated settlement is attainable. This article addresses these theoretical and empirical gaps through three contributions. First, it refines rationalist bargaining approaches by identifying commitment challenges that explain why conflict actors rationally avoid both fighting and sincere negotiations, sustaining non-decisive equilibria such as mutually beneficial stalemates. Second, it develops a typology conceptualizing three distinct categories of non-decisive conflict outcomes—Territorial Change, Rebel Reboot, and Conflict Management—advancing explanations for previously overlooked empirical cases. Third, it illustrates variation across non-decisive outcomes through an empirical analysis of Myanmar's armed conflicts. The findings demonstrate that non-decisive outcomes function as strategic pathways shaped by rational choices, helping scholars and policymakers more effectively recognize and study these alternative endpoints.

The "AI race" or "AI arms race" refers to the intensifying competition, primarily between the United States and China, for advantage in artificial intelligence. This article argues that the contest comprises four overlapping domains of competition, each with its own dynamics. The first is frontier AI model development: who builds the most advanced models and at what cost. The second is AI infrastructure: who controls the inputs needed to develop and train advanced AI, including critical minerals, chips, compute, data, and energy. The third is national AI diffusion: how AI is applied to support national security goals, gain military advantage, and advance productivity and economic growth. The fourth is international AI diffusion: who supplies AI capability and infrastructure to other countries, on what terms, and with what effect on ecosystem alignment. Disaggregating the "AI race" helps clarify where the US and China compete and where they pursue different goals. Turning to the AI race narrative itself, the article contends that by implying a finish line and single winner, the framing obscures what is likely long-term strategic competition with limited periods of technological advantage, multi-decade AI infrastructure policies, and increasingly different AI ecosystems. It concludes that assessing each area of competition separately helps identify where shared risks make cooperation possible, and that emerging confidence-building measures on AI safety offer a modest start.

This article introduces the Intrastate Conflict Termination and Outcomes (ICTO) dataset covering conflict endings between 1946 and 2023. Civil war termination is more than violence stopping. It is a bargaining process with both sides accepting the symbols of defeat or victory, or the terms of peace. When settlements or winning are out of reach, belligerents often turn to conflict management strategies, maintaining ceasefires and mutually beneficial stalemates. The transformations of conflict actors also shape short- and long-term conflict ending processes, including territorial changes due to state collapse or chronic failure, with rebel groups carving out durable statelets. Multiple rebel groups may merge institutions, thereby ending one conflict and starting a new dyad. Rebel groups may also surrender or disband long after violence has stopped, giving governments a gradual victory off the battlefield. Many of these endings fall outside the window of existing termination datasets, which code outcomes from the last year of fighting and the first year of inactivity. The ICTO addresses this gap by tracing conflict actors and their endpoints five or more years after violence subsides, accounting for conflict outcomes which occur in periods of non-violence. The dataset also introduces theorized categories of non-decisive conflict outcomes, the most frequent scenarios, which lack the obvious symbolic cues of a peace agreement or military outcome. The sharpened conceptualization of conflict termination and outcomes, combined with the tracing approach, produces a rise in government victory and peace agreements with the decline in non-decisive situations.

Multi-actor conflicts have become the dominant form of civil war, affecting 65 percent of conflict countries during 2015-2024. I examine how rebel proliferation affects outcomes by analyzing intrastate conflicts from 1946-2012 using random-effects multinomial logit models. Multi-actor dynamics reduce military victories and peace agreements while producing conflict-management outcomes involving reduced violence and limited cooperation. Rebel splintering constrains government victory (89%), whereas concurrent original rebels reduce both government (76%) and rebel victories (90%). When settlements occur in fragmented conflicts, 83 percent are partial agreements excluding groups, highlighting sequential peacemaking pathways. These findings support my theoretical argument: conflict parties strategically pursue non-decisive outcomes sustained by mutually beneficial stalemates—equilibria in which maintaining the status quo yields higher returns than renewed fighting for uncertain military gains or comprehensive settlements that risk commitment failures. In multi-actor settings, rebels and states pursue organizational survival, with states avoiding overextension and rebels limiting violence through conflict-management scenarios.