Global Peace Index Data: AI Making Life-and-Death Combat Decisions as Global Conflict Hits Record High

The global security environment is undergoing its most profound transformation since the end of the Second World War, driven by a record number of active conflicts and the rapid integration of artificial intelligence into lethal decision-making. According to the 2026 Global Peace Index, released by the Institute for Economics & Peace, global peacefulness has deteriorated for the twelfth consecutive year. The index recorded 61 active state-based conflicts, the highest number since 1945, alongside a six-fold increase in internal conflict deaths since 2007. This structural decline is unfolding against the backdrop of what researchers term the "Great Fragmentation," a geopolitical shift where traditional powers wane and middle powers assert influence. However, the most disruptive force identified in the data is not purely political, but technological: the delegation of life-and-death combat decisions to machine-speed algorithms.[1]

For the first time in the history of warfare, autonomous systems and artificial intelligence are compressing the "kill chain"—the sequence of events from identifying a target to engaging it—faster than human operators can meaningfully review. In the 1990s, the targeting cycle for a cruise missile strike typically took around 24 hours of human analysis, legal review, and command authorization. Today, algorithmic targeting systems have reduced that window to a matter of seconds. This velocity gap between human cognition and machine computation is fundamentally altering military doctrine, forcing armed forces to choose between maintaining strict human oversight and keeping pace with adversaries who deploy fully autonomous weapons.[1]

The mechanical reality of this shift is starkly visible in recent conflict zones. In Gaza, the deployment of algorithmic targeting systems has reportedly compressed the human review of AI-generated targets to roughly 20 seconds per strike. At this speed, human operators are largely relegated to rubber-stamping machine outputs rather than conducting independent verification of a target's validity or assessing the proportionality of the strike. This near-instantaneous execution removes the friction that traditionally slowed the pace of warfare, allowing for an unprecedented volume of engagements but severely eroding the capacity for moral judgment and legal compliance under international humanitarian law.[1]

Parallel to the compression of decision time is the exponential proliferation of the delivery mechanisms themselves. The 2026 Global Peace Index documents an 11,500 percent increase in recorded drone attacks between 2018 and 2025. This surge is not confined to advanced nation-states; the data reveals that 565 different armed groups, including non-state actors and criminal cartels, carried out drone strikes during this period. The convergence of cheap, accessible drone hardware with increasingly sophisticated AI software has democratized lethality, lowering the financial and technical barriers to precision warfare and allowing smaller groups to project power previously reserved for superpowers.[1]

In highly contested environments, the operational reality has pushed autonomy even further, effectively killing the traditional requirement for a "human-in-the-loop." Modern front lines are characterized by drone saturation zones where intense electronic warfare and signal jamming routinely sever the communication links between weapons and their human operators. To ensure munitions reach their targets despite this interference, defense contractors have developed systems like the Long Range Anti-Ship Missile (LRASM), which features autonomous targeting capabilities designed to detect and destroy specific targets without a continuous data link. When the human connection is jammed, the machine relies entirely on its onboard AI to select and strike.[1]

This battlefield reality exposes a critical flaw in traditional military doctrine. Conventional targeting methodologies are structured around sequential human validation, assuming that command authority is exercised close to execution and that a human will validate every step. However, as sensing and striking converge into a single, near-simultaneous machine process, this doctrine concentrates human judgment precisely at the point where time is least available. The result is a doctrinal bottleneck that prevents autonomous systems from operating at their designed pace, reintroducing latency and leaving forces vulnerable to faster, fully autonomous adversaries.[1][2]

To navigate this tension, defense analysts and legal scholars categorize weapons along an "autonomy spectrum." At one end are semi-autonomous systems, or "human-in-the-loop" weapons, which require an operator's affirmative action to use lethal force against a specific target. In the middle are "human-on-the-loop" systems, where the machine selects and engages targets independently, but a human supervisor monitors the operation and can intervene to abort the strike. At the far end are fully autonomous weapons, which, once activated, execute their programming without any further human intervention or possibility of recall.[2]

The shift from human-in-the-loop to human-on-the-loop is not merely a technical upgrade; it is a fundamental relocation of human responsibility. Proponents argue that delegating tactical execution to machines can actually save lives by removing human error, fatigue, and panic from the equation. In highly contested environments, commanders are moving away from approving individual targeting episodes. Instead, human judgment is exercised earlier in the process through program selection and geographic bounding. A commander authorizes an autonomous system to engage specific classes of targets within a strictly defined geographic box and time window, accepting the associated risks while the machine handles execution at superhuman speed.[2][3]

While this approach closes the velocity gap, it introduces profound legal and ethical challenges. International humanitarian law, specifically the Geneva Conventions, requires armed forces to distinguish between combatants and civilians and to ensure that any incidental civilian harm is proportional to the direct military advantage anticipated. These principles require nuanced, context-specific moral reasoning. Translating the concepts of distinction and proportionality into rigid algorithmic code remains an unsolved technical challenge, raising alarms among humanitarian organizations about the inability of machines to exercise restraint or recognize surrender.[2]

The accountability gap further complicates the deployment of lethal autonomous weapons systems. When a human soldier commits a war crime, the chain of command and international tribunals provide a mechanism for investigation and prosecution. However, when an AI system misidentifies a civilian convoy as a military convoy due to algorithmic bias or a sensor malfunction, attributing legal responsibility becomes murky. Is the commander who deployed the system liable? The software engineer who wrote the code? The procurement officer who approved the purchase? Existing legal frameworks struggle to assign culpability for the actions of inanimate objects operating independently.[2]

Despite these unresolved ethical dilemmas, the economic incentives driving the adoption of AI in warfare are overwhelming. The global economic impact of violence reached a staggering $21.81 trillion in 2025, equivalent to 10.5 percent of global GDP. Concurrently, global military expenditure hit a record $2.9 trillion. Autonomous systems offer armed forces a way to project power more cheaply and efficiently, reducing the need for expensive human personnel and minimizing friendly casualties. As the marginal cost of lethality drops, the strategic calculus shifts heavily in favor of mass-produced, AI-enabled drone swarms over traditional, human-crewed platforms.[1][3]

International governance has entirely failed to keep pace with this technological revolution. While activists and dozens of nations have campaigned for a global treaty banning lethal autonomous weapons systems, the window for a preemptive ban has likely closed. The technology is already deployed, proven, and rapidly proliferating. Furthermore, 118 of the 193 United Nations member states are absent from key agreements regarding autonomous weapons, highlighting a severe lack of global consensus. Major military powers have consistently resisted binding regulations that would constrain their ability to develop and field AI-enabled systems, preferring flexible, non-binding guidelines.[1][2]

The lack of a unified regulatory framework means that the rules of engagement for autonomous weapons are being written in real-time on the battlefields of Eastern Europe and the Middle East, rather than in diplomatic chambers in Geneva. This ad hoc approach increases the risk of unintended escalation. Research indicates that human operators are highly susceptible to "automation bias," a psychological phenomenon where individuals tend to defer to computer-generated decisions over their own perceptions. In a crisis, an operator is highly likely to approve an AI system's recommendation to use force, potentially triggering a rapid, machine-driven escalation before diplomatic channels can be opened.[1][2]

The integration of AI into combat is also reshaping the physical infrastructure of warfare. The massive computational power required to train and run frontier AI models demands vast amounts of energy. The 2026 Global Peace Index notes that data center electricity use is projected to reach 945 terawatt-hours by 2030, doubling from 2024 levels. This physical footprint makes energy grids and data centers critical military targets, further expanding the scope of modern conflict and intertwining civilian infrastructure with national security imperatives.[1]

An analysis of the 2026 data reveals the stark mechanical relationship between decision speed and strike volume. By comparing the historical 24-hour targeting cycle to the modern 20-second algorithmic review, it becomes evident that the lethal decision-making window has been compressed by 99.97 percent. This near-total elimination of human friction correlates directly with the 11,500 percent exponential surge in drone strike volume over the past seven years. The data demonstrates that removing the human-in-the-loop is not merely a side effect of modern warfare, but the primary mechanical driver enabling the sheer scale and frequency of autonomous engagements today.[1][4]

Ultimately, the transition to AI-augmented warfare represents a point of no return for global security. The velocity of modern combat has outstripped the biological limits of human cognition, making autonomous systems a tactical necessity for any advanced military. As the 2026 data clearly demonstrates, the friction of human review is being systematically engineered out of the kill chain. The challenge for the next decade will not be preventing the rise of autonomous weapons, but rather developing new doctrines and legal frameworks capable of imposing meaningful boundaries on machines that decide who lives and who dies in fractions of a second.[1][2][3][4]