The Ghost in the War Room: How AI is Marginalizing Security Studies
For three-quarters of a century, the study of international security was rooted in a foundational premise: war and peace are human endeavors. From the nuclear brinkmanship of the Cold War to the counterinsurgency doctrines of the post-9/11 era, Security Studies as a discipline prospered by deciphering human psychology, statecraft, institutional incentives, and strategic signaling. Scholars spent decades refining theories of deterrence, off-the-shelf bargaining models, and balance-of-power dynamics. All grounded in the assumption that leaders subject to fear, honor, and interest ultimately weighed decisions as reflected in the work of Thomas Schelling and James Fearon.
Today, that intellectual superstructure is facing an existential crisis. As artificial intelligence moves from the realm of tactical automation to strategic decision-support, predictive battle management, and autonomous systems, traditional Security Studies is not merely being challenged—it is being marginalized.
AI is creating a rupture in how strategic knowledge is produced and used. As algorithms assume the burden of processing intelligence, mapping escalation pathways, and targeting at speeds measured in milliseconds, the traditional tools of security analysis are proving dangerously slow and conceptually obsolete. If security scholars do not fundamentally overhaul their analytical toolkits, the discipline risks degenerating into a historical archive, ceding the future of strategic thought entirely to computer engineers, data scientists, and defense contractors.
The Algorithmic Eclipse of Human Agency
The core of Security Studies’ marginalization lies in a shift of agency. Classic security literature—whether Realist, Liberal, or Constructivist—assumes a human sovereign who perceives a threat, processes information through institutional channels, and acts rationally or semi-rationally.
Artificial intelligence subverts this chain of logic in three distinct ways:
- The Compression of Decision Time: Traditional theories of deterrence rely on communication, intent, and deliberative pause. Under “hyperwar”—where AI systems orchestrate drone swarms, cyber counter-offensives, and electronic warfare simultaneously—the human decision-making loop is collapsed. When strategic success hinges on microsecond response times, human deliberation becomes a vulnerability rather than a safeguard.
- The “Black Box” of Machine Epistemology: Classical security analysis relies on interpreting the intent of an adversary. But advanced deep-learning systems operate through non-linear multi-layered neural networks whose internal logic is opaque even to their creators. How do you analyze an adversary’s “strategic culture” or “doctrine” when their battlefield deployments are driven by an uninterpretable algorithmic optimization curve?
- The Privatization of Strategic Insight: Historically, academia and policy think tanks held a monopoly on strategic conceptualization. Today, the cutting edge of military AI technology is held inside private tech giants and secretive defense-tech startups. The primary architects of modern security are no longer tenure-track professors or diplomats; they are Silicon Valley software developers who rarely read Clausewitz, Schelling, or Jervis.
A Discipline Trapped in Analog Concepts
Security Studies has also contributed to its own vulnerability by treating AI as another weapons technology rather than as a challenge to the way strategy, itself is understood. For too long, mainstream academic journals have treated AI as merely another “weapon system” to be plugged into existing theoretical paradigms—much like the tank or the radar.
This is a fundamental misdiagnosis. AI is not merely a tool within security; it is a meta-technology that fundamentally alters what security means.
Applying 20th-century strategic concepts to algorithmic warfare creates dangerous intellectual blind spots:
- Deterrence Theory Is Fraying: Deterrence requires clear attribution, credible threats, and mutual understanding of thresholds. Algorithmic assets—spanning autonomous cyber exploits, deepfake-driven cognitive warfare, and uncrewed autonomous platforms—are inherently designed for ambiguity, deniability, and continuous, sub-threshold friction.
- Escalation Dominance Is Unpredictable: When autonomous AI systems interact on the battlefield, their unexpected interaction effects—known as “flash cascades”—can trigger unintended escalations before human commanders even realize a threshold has been breached. Traditional escalation ladders simply cannot capture machine-speed feedback loops.
The Risk: Tech-Determinism Without Strategic Wisdom
The marginalization of security scholars has created a dangerous vacuum. As political scientists and strategists are sidelined, global security, policy is increasingly driven by a naive technological determinism.
Silicon Valley’s ethos—move fast and break things—is catastrophically ill-suited for nuclear-armed international systems. Computer scientists often treat security as an engineering problem to be solved through data optimization, overlooking the historical, ideological, and socio-political drivers of conflict. An algorithm can optimize target acquisition, but it cannot comprehend the historical memory that drives a nation to endure catastrophic loss rather than surrender.
An algorithm can optimise target acquisition, but it cannot comprehend the historical memory that drives a nation to endure catastrophic loss rather than surrender.“
Without the critical, qualitative, and historical depth that Security Studies uniquely provides, the international community risks building automated security architectures that are technically brilliant yet strategically catastrophic.
Reclaiming the Discipline: A Blueprint for Survival
If Security Studies is to regain its relevance, it must undergo a radical metamorphosis. Scholars cannot content themselves with writing post-hoc critiques of technologies they do not understand.
- Epistemic Hybridity: Security scholars must achieve technological literacy. Future curricula in security studies must blend international relations theory with data science, machine learning architectures, and system dynamics.
- Focus on Human-Machine Teaming: Rather than studying human decisions or machine outputs in isolation, the field must pioneer theories of hybrid agency—understanding how human commanders interact with, rely on, or misinterpret AI-generated advice under stress.
- Pioneering Algorithmic Arms Control: Traditional arms control treaties relied on counting warheads or inspecting silos. AI code cannot be inspected in the same way. Security scholars must work alongside computer scientists to invent new paradigms of verification, normative frameworks, and confidence-building measures for software-defined power.
The tragedy of contemporary strategy is that at the exact moment we need profound human judgment most, we are delegating the cognitive burden of security to machines.
Security Studies was born in the shadow of the atomic bomb to answer a fundamental human question: How do we survive our own inventions? Today, as we stand on the threshold of an era defined by autonomous violence and machine-calculated strategy, that question is more urgent than ever.
If security scholars allow themselves to be pushed to the margins of this technological revolution, they will not just be forfeiting their academic discipline—they will be abandoning the human element in the ultimate decisions of life, death, and global survival.
(The opinion expressed in this article belong only to the author and do not necessarily reflect the views of The Convention).
