Artificial Intelligence and International Peace and Security: SIPRI View

Representative Image by Wix AI

The Stockholm International Peace Research Institute's (SIPRI) Yearbook 2026 has a chapter, “Artificial Intelligence and International Peace and Security,” authored by Laura Bruun, Jules Palayer and Vincent Boulanin.

The authors offer a comprehensive and carefully sourced account of how the international community grappled with AI governance in 2025.

Rather than focusing solely on technological developments, the report examines how governments, international organizations, and multilateral institutions are attempting to regulate AI's military and civilian applications. It effectively highlights that AI governance is no longer a theoretical concern but a pressing international security issue requiring coordinated global action.

One of the report's greatest strengths is its structured approach. It divides the discussion into two broad domains: governance of military AI and governance of civilian AI. Military AI raises concerns regarding the use of force, accountability, escalation, and compliance with international humanitarian law, whereas civilian AI governance focuses on issues such as digital inequality, transparency, safety, and responsible innovation. This separation makes the report have analytical depth.

The treatment of autonomous weapon systems (AWS) is the chapter's most important section. The Chapter demonstrates that after more than a decade of negotiations under the UN Convention on Certain Conventional Weapons (CCW), states continue to disagree on even the most fundamental concepts, including the definition of AWS and the level of human involvement necessary in targeting decisions. The report carefully explains the growing support for a two-tier regulatory approach that would prohibit certain autonomous weapons while regulating others according to international humanitarian law. However, it also reveals that political disagreements between major military powers continue to prevent meaningful progress.

A major contribution of the report is its discussion of "context-appropriate human control and judgment." Rather than insisting on the notions of "human in the loop," the authors explain that meaningful human oversight depends on the operational context. This reflects a more realistic understanding of modern warfare, where AI increasingly supports intelligence analysis, surveillance, logistics, and decision-making rather than replacing human operators. At the same time, the report emphasizes that responsibility and accountability for the use of force must always remain with humans, reinforcing an essential principle of international humanitarian law.

Another important aspect highlighted by the authors is the expansion of AI governance discussions beyond Geneva. Traditionally, debates on autonomous weapons have taken place under the CCW framework. During 2025, however, the United Nations General Assembly in New York hosted informal consultations on AWS, while the Security Council also debated AI's implications for international peace and security. This geographical and institutional expansion illustrates that AI governance is evolving into a broader international security issue rather than remaining confined to arms control discussions. SIPRI effectively argues that while Geneva remains the primary negotiating forum, New York offers greater inclusivity and political visibility.

The Chapter also provides valuable analysis of the UN Secretary-General's first report on AI in the military domain. Instead of limiting the discussion to autonomous weapons, the report examines AI applications in intelligence, surveillance, logistics, cybersecurity, command and control, peacekeeping, and strategic analysis.

The authors identify both opportunities and risks. On one hand, AI can improve operational efficiency, precision, and civilian protection. On the other hand, it may shorten decision-making timelines, increase the risk of accidental escalation, widen technological inequalities between states, and reduce meaningful human control over military operations. This balanced assessment avoids both technological optimism and excessive alarmism.

Beyond military applications, SIPRI successfully broadens the conversation to civilian AI governance. The report discusses significant United Nations initiatives, including the establishment of the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance. These initiatives demonstrate growing recognition that AI governance requires scientific expertise alongside diplomatic negotiations. The emphasis on inclusivity and representation from both developed and developing countries is significant, as AI governance risks becoming dominated by technologically advanced states and large private corporations.

The chapter also offers a detailed examination of major international AI governance initiatives outside the United Nations. The AI Action Summit in Paris illustrates the shift from focusing exclusively on AI safety toward broader discussions on sustainability, inclusivity, and economic development. Similarly, SIPRI reviews the European Union's implementation of the AI Act, China's Global AI Governance Action Plan, and the United States' AI Action Plan under the Trump administration. These comparisons reveal that AI governance is increasingly shaped by geopolitical competition rather than purely technical considerations.

Perhaps the author’s most significant contribution is its demonstration of growing divergence among major powers. The European Union emphasizes regulation, transparency, and responsible innovation. China promotes multilateral cooperation while simultaneously expanding its own technological influence through capacity-building initiatives. In contrast, the United States increasingly prioritizes technological leadership, deregulation, and strategic competition with China. SIPRI effectively illustrates that AI governance is becoming another arena of great-power rivalry, making the achievement of universal international norms increasingly difficult.

Limitations

First, emphasis remains heavily institutional and geopolitical. While the report thoroughly documents negotiations, resolutions, and international meetings, it devotes comparatively less attention to how AI is already transforming contemporary battlefields. Practical case studies from ongoing conflicts could have strengthened the analysis by illustrating the real-world consequences of AI-enabled military systems.

Second, the report focuses primarily on state-led governance. Although it acknowledges the influence of major technology companies, the discussion could have explored in greater depth the role of private industry. Companies such as OpenAI, Google, Microsoft, Anthropic, Deep Seek and others are driving much of AI innovation, yet international governance mechanisms still struggle to regulate actors whose capabilities often rival those of governments. Greater analysis of public-private governance would have enhanced the report.

Third, while SIPRI recognizes the importance of capacity-building for developing countries, the report pays relatively limited attention to the challenges faced by the Global South. Many developing countries lack the technical expertise, computational infrastructure, and regulatory capacity required to participate effectively in global AI governance. Future editions could examine these disparities more comprehensively.

The chapter notes that IISPAI has a 40-member, geographically balanced design and that the Global Dialogue aims to give “every country a seat at the table.” However, it does not examine whether this formal inclusion leads to meaningful influence in decision-making. The chapter also highlights that the UN Secretary-General received only 49 submissions on military AI, compared with 73 submissions on LAWS the previous year. It attributes this decline to states having “less developed positions.” This presented an opportunity to explore whether differences in technical capacity, rather than representation alone, affect which states are able to shape these discussions. While the chapter's own data points to this issue, it does not examine it further.

Nevertheless, these limitations do not significantly diminish the report's overall value. SIPRI succeeds in presenting an objective, evidence-based assessment without advocating a single regulatory model. Instead, it carefully documents competing perspectives, allowing readers to understand the complex political, legal, and technological dimensions of AI governance.

Conclusion

As an overview of AI governance in 2025, this chapter is well-researched, and clearly explains where states disagree. Its discussion of the CCW's definitional deadlock and the debate over “context-appropriate human control and judgment” is one of the strongest parts of the chapter. The comparison of the EU, China, and US positions is also clear and well-supported.

The chapter's main limitation is its scope rather than its accuracy. It focuses on diplomatic processes and institutional developments but gives less attention to broader questions, such as how AI systems perform in practice, how corporate influence shapes governance, and whether inclusive participation leads to meaningful influence. Overall, the chapter is a valuable resource providing a clear account of AI governance developments in 2025. However, while it presents the evidence effectively, it stops short of assessing whether these diplomatic efforts are producing meaningful progress.

Note - Stuti Sharma is a research intern with Security Risks Asia