Loading module…
Loading module…
Module 1 of 6
Artificial intelligence is reshaping national security and warfare as profoundly as any technology in generations, and understanding this transformation is essential — for analysts, for policymakers, and for citizens who must think clearly about one of the defining strategic issues of the era. This course examines AI in warfare and national security at the analytical, strategic, and policy level: what military AI actually is and does, the strategic dynamics it creates, the genuine risks it poses, and the governance and ethical debates it forces. It is a course for understanding a critical topic clearly and honestly — not an operational manual, but an analytical education in one of the most consequential and contested applications of the technology whose foundations the AI courses established.
The frame for the whole course is analytical and honest, and two commitments govern it. First, this is treatment at the level of understanding — the strategic dynamics, the capabilities and their implications, the risks, and the governance questions — as a war-studies or security-policy course would treat any major military-technological development, not operational detail for building or employing systems. Second, honesty over both hype and alarm: military AI is surrounded by breathless claims (of revolution) and dire warnings (of catastrophe), and this course, committed to factual accuracy, distinguishes what is real and demonstrated from what is speculative or exaggerated, treating the genuine capabilities, the genuine risks, and the genuine uncertainties clearly rather than adopting either the techno-optimist or the doomsayer distortion. From that honest, analytical ground, the course builds a clear understanding of AI's role in national security — its capabilities, dynamics, risks, and governance — which is what informed analysis of this critical topic requires.
To understand AI in national security you must first understand why it matters so much — why AI is treated as a transformative, even decisive, technology by defense establishments worldwide — because the reasons explain the intensity of the competition and the stakes of the governance debates. AI matters to national security for structural reasons rooted in what the technology does.
The foundational reason is that AI is a general-purpose technology that enhances a vast range of military and security functions, not a single weapon but a capability that improves nearly everything it touches. Just as electricity or computing transformed warfare not as a single weapon but as an enabling technology across all functions, AI enhances intelligence (processing vast data — Module 3), command and decision-making (analysis and speed), logistics and operations (optimization), cyber capabilities (Module 4), autonomous systems (Module 2), and more. This general-purpose, cross-cutting character is why AI is seen as strategically decisive: a state that leads in AI could enhance its military and security across many dimensions simultaneously, a broad advantage rather than a single-weapon edge. The breadth is the point — AI is a force multiplier across the whole national-security enterprise, which is why it is pursued so intensely and why leadership in it is treated as a strategic imperative.
A second reason is speed and scale — AI enables processing, decision, and action at speeds and scales beyond human capacity, which in security contexts can be decisive. AI can process intelligence data, detect patterns, and support decisions far faster than humans (Module 3); it can operate at scales (many systems, vast data) humans cannot match; and in some contexts, the speed advantage is itself strategically significant (faster detection, faster decision, faster response). This speed-and-scale enhancement is particularly consequential in security, where speed can determine outcomes and where the side that can decide and act faster may prevail — creating pressure to adopt AI for the speed advantage, with attendant risks (Module 5) when speed outruns human judgment. The speed-and-scale character of AI is thus both a source of its security value and a source of its risks, and it is central to why AI matters strategically.
A third reason is the competitive dynamic — because AI is strategically valuable, states compete to lead in it, and this competition itself makes AI matter more, creating a dynamic where no major power can afford to fall behind. The perception that AI is strategically decisive drives an AI competition among major powers, each seeking to lead (or not lag) in military AI, and this competition intensifies the stakes: AI matters not just for its direct capabilities but because the competition to lead in it is itself a major strategic dynamic, with the leaders potentially gaining significant advantage and the laggards potentially falling dangerously behind. This competitive dynamic (developed in Section 4) is a large part of why AI matters to national security — it is not just a useful technology but the object of a strategic competition that shapes the security environment. Together these reasons — AI as a general-purpose enhancer across security functions, its speed-and-scale advantages, and the competitive dynamic its value creates — explain why AI is treated as a transformative national-security technology, why it is pursued so intensely, and why the stakes (and thus the governance debates) are so high. Understanding why AI matters is the foundation for understanding everything the course examines: the capabilities, the dynamics, the risks, and the governance all flow from AI's structural significance to national security.
Key Insight: AI matters to national security for structural reasons: it's a general-purpose technology enhancing a vast range of functions (intelligence, decision-making, logistics, cyber, autonomous systems) rather than a single weapon — a broad, cross-cutting force multiplier, which is why leadership in it is treated as strategically decisive; it enables speed and scale beyond human capacity, which can determine outcomes (and creates risks when speed outruns judgment); and its strategic value drives a competitive dynamic among major powers where none can afford to lag, itself a major strategic force. These structural reasons explain the intensity of pursuit and the height of the stakes.
Understanding military AI requires a map of its actual applications — the range of ways AI is being applied across the national-security enterprise — because the reality is a broad landscape of diverse applications, most of them far from the autonomous-weapons image that dominates popular attention. This section surveys the landscape at the analytical level, establishing what military AI actually comprises.
The largest and least dramatic category is AI for information processing and decision support — using AI to process the vast data of modern intelligence and operations, and to support (not replace) human decision-making. This includes intelligence processing (Module 3 — analyzing imagery, signals, and vast data at scale), pattern detection and warning, logistics and resource optimization, predictive maintenance, and decision-support tools that help commanders and analysts by processing information and surfacing insights. This category is where most military AI actually is — enhancing the information and decision processes of security, largely in support roles — and it is the least controversial and most established, extending the information-processing enhancement AI provides everywhere into the security domain. It is important to recognize that this large, mundane category (AI processing information to support human decisions) is the bulk of military AI, not the autonomous weapons that dominate attention — the reality is more about better intelligence-processing and decision-support than about killer robots.
A second category is autonomous and semi-autonomous systems — the drones, vehicles, and platforms that operate with varying degrees of autonomy, and the weapons systems that raise the autonomous-weapons debate (Module 2). This ranges from remotely-operated systems (human-controlled) through semi-autonomous (some functions automated, human in control) to the debated fully-autonomous systems (able to select and engage without human intervention). This category includes the most controversial military AI — the autonomous weapons that Module 2 examines in depth and that dominate the ethical debate — but it also includes much that is less controversial (autonomous logistics, reconnaissance, non-weapon systems). The autonomy dimension (how much a system does without human control) is the key variable, and it ranges widely, with the fully-autonomous-weapons end being the most contested. A third category is cyber and information operations — AI applied to cyber capabilities (Module 4) and to the information domain (influence, disinformation detection and generation), where AI enhances both offensive and defensive cyber and information capabilities, raising distinct issues the relevant modules examine.
Further categories round out the landscape. Command, control, and coordination — AI enhancing how forces are commanded and coordinated, potentially at machine speed, which raises the strategic-risk questions of Module 5 (speed, escalation, human control). Training and simulation — AI for realistic training and wargaming. Administrative and enterprise — the mundane AI applications (like any organization's) in the vast defense enterprise. Across this landscape, the analytical points are: military AI is a broad and diverse field, most of it far from the autonomous-weapons image; the applications range from mundane (information processing, logistics, administration) to profound (autonomous weapons, machine-speed command); and the autonomy dimension (how much AI does without human control) is the key variable distinguishing the more-established from the more-contested applications. Understanding this landscape — the range of military AI applications, the predominance of the information-processing-and-support category, and the autonomy dimension that distinguishes the contested applications — is what grounds analysis of military AI in reality, correcting the popular focus on autonomous weapons alone and revealing the broad, diverse field that military AI actually is. The subsequent modules examine the most significant and contested categories (autonomous weapons, intelligence processing, cyber operations) in depth, but this landscape shows their place in the larger field, most of which is the less-dramatic enhancement of information and decision processes.
Key Insight: Military AI is a broad, diverse landscape, most of it far from the autonomous-weapons image: the largest category is AI for information processing and decision support (intelligence analysis, pattern detection, logistics, decision-support tools — largely supporting human decisions, the established bulk of military AI); then autonomous and semi-autonomous systems (ranging from remotely-operated through semi-autonomous to the contested fully-autonomous weapons — the autonomy dimension being the key variable); cyber and information operations; command and coordination; training; and mundane enterprise applications. The reality is more about better intelligence-processing and decision-support than killer robots — and the autonomy dimension distinguishes the established from the contested.
An honest analysis of military AI requires being clear about what it can and cannot actually do — because military AI, even more than commercial AI, is surrounded by hype (of revolutionary capability) and by fear (of superhuman autonomous systems), and the honest capability picture, grounded in the AI courses' understanding, corrects both. This section applies the AI courses' honest capability assessment to the military domain.
Military AI's genuine capabilities are real and significant, and they are the capabilities of the AI the earlier courses described — powerful pattern recognition, information processing at scale, and automation of definable tasks. AI genuinely can process vast intelligence data and detect patterns humans would miss (Module 3), operate systems with real autonomy in defined conditions, enhance cyber capabilities (Module 4), and support decisions with rapid analysis — these are real, demonstrated capabilities that provide genuine military value, and they are why AI matters (Section 1). The enhancement of information processing, the automation of definable functions, and the pattern-recognition-at-scale are genuine and consequential military capabilities, not hype — military AI really does provide these, and they really do enhance security functions significantly.
But military AI has exactly the limits of AI generally (the AI courses' failure modes), and these limits are especially consequential in the military context where the stakes of failure are extreme. Brittleness: military AI, like all AI, is brittle — reliable within its training distribution and prone to fail on the novel, unexpected situations that warfare, adversarial and chaotic, constantly produces. The adversarial nature of war makes this acute: adversaries will deliberately produce the out-of-distribution situations and adversarial inputs that fool AI (the adversarial examples of the AI courses, weaponized), so military AI faces an environment specifically hostile to its reliability. Lack of judgment and understanding: military AI has the AI courses' lack of genuine understanding, common sense, and judgment — it pattern-matches without comprehending, which in the military context means it can fail to grasp the situation, misjudge the novel case, and lack the contextual judgment that military decisions (especially consequential ones) require. Unreliability and unpredictability: military AI has AI's characteristic unreliability and can fail unpredictably, which in high-stakes military contexts (where failures can be catastrophic) is especially dangerous. These limits mean military AI is powerful for the definable, in-distribution tasks and unreliable for the novel, adversarial, judgment-requiring situations that war abounds in — a crucial honest limit.
The honest capability picture thus corrects both distortions and grounds sound analysis. Against the hype (military AI as revolutionary, superhuman, decisive): military AI is powerful but limited, brittle in the adversarial chaos of war, lacking judgment, and unreliable in the novel situations warfare produces — not the reliable superhuman capability the hype suggests. Against the fear (autonomous AI as unstoppable superhuman force): military AI is the limited, brittle, judgment-lacking AI of the earlier courses, not a superintelligence — its dangers (Module 5) come more from its unreliability and the risks of relying on it than from superhuman capability. The honest picture: military AI provides genuine, significant capabilities (information processing, automation, pattern recognition at scale) and has genuine, significant limits (brittleness, no judgment, unreliability) especially consequential in war's adversarial, high-stakes, novel-situation-rich environment. This honest picture — real capabilities, real limits, especially the limits' danger in the military context — is what grounds sound analysis of military AI, neither hyping its power nor fearing a superhuman capability it does not have, and it frames the risks (Module 5) as flowing largely from the interaction of AI's real limits with war's demanding environment and the pressure to rely on AI despite those limits. Analyzing military AI honestly means holding both its genuine power and its genuine limits, especially the limits that make relying on it in war's novel, adversarial, high-stakes situations so risky.
Key Insight: Military AI has the genuine capabilities of AI generally — powerful pattern recognition, information processing at scale, automation of definable tasks — which provide real, significant military value. But it has AI's limits, especially consequential in war: brittleness (unreliable outside its training distribution — acute because adversaries deliberately produce the out-of-distribution and adversarial inputs that fool it), lack of judgment and understanding (pattern-matching without comprehending, failing the novel case war abounds in), and unreliability (dangerous where failures are catastrophic). The honest picture corrects both hype (it's powerful but limited and brittle, not revolutionary/superhuman) and fear (it's the limited AI of the earlier courses, not superintelligence) — and frames the risks as flowing from AI's real limits meeting war's demanding environment.
The competition among major powers to lead in military AI is one of the defining strategic dynamics of the era, and understanding it — its drivers, its character, and its dangers — is essential to understanding AI's role in national security, because the competition shapes how AI is developed and deployed and creates risks of its own. This section analyzes the strategic competition at the level of its dynamics.
The competition is driven by the perception that AI leadership confers decisive strategic advantage (Section 1) — since AI is seen as a general-purpose enhancer across military functions with potentially decisive speed-and-scale advantages, states perceive that leading in AI could confer significant strategic advantage and that lagging could mean falling dangerously behind. This perception drives major powers to invest heavily in military AI and to compete for leadership, creating an AI competition analogous to previous strategic-technology competitions. The competition's intensity flows from the stakes (perceived decisive advantage) and from its self-reinforcing character — because others are competing, none can afford not to, so the competition sustains and intensifies itself, each power's investment justifying the others'. This dynamic — heavy investment and competition driven by the perception of decisive stakes and sustained by its self-reinforcing character — is the basic shape of the military-AI competition.
The competition has a distinctive and consequential character that raises specific dangers. The pace pressure: the competition creates pressure to develop and deploy military AI quickly, to not fall behind — and this pace pressure is dangerous because it can push development and deployment faster than the reliability, testing, and governance the technology's risks require (Module 5), trading safety for speed in the competitive rush. The pressure to deploy quickly, lest an adversary gain advantage, works against the caution AI's limits and risks warrant, which is a central danger of the competition. The dual-use and diffusion character: military AI builds heavily on commercial AI (the same underlying technology), so it diffuses more readily than some military technologies (the capabilities are broadly available, not tightly controlled), which shapes the competition (many actors can access military-relevant AI) and complicates control. The uncertainty: because AI's military value is real but its limits are significant (Section 3), the competition proceeds under uncertainty about how decisive AI actually is, which can drive both over-investment (in case it is decisive) and miscalculation (about relative capabilities).
The competition's dangers are a major theme the course develops, and they are worth previewing here as part of the strategic dynamics. The pace-versus-safety tension (deploying faster than safety warrants, driven by competition) is a central risk (Module 5). The escalation and stability risks (AI affecting crisis stability, escalation, and the speed of conflict — Module 5) are shaped by the competition. The governance challenge (Module 6) — the difficulty of governing military AI when competition pushes against restraint and when the technology diffuses readily — is central. And the arms-race dynamics — the risk that the competition becomes a destabilizing arms race, with the pace pressure and mutual investment driving rapid, insufficiently-governed militarization of AI — is a defining concern. Understanding the strategic competition — its drivers (perceived decisive stakes), its self-reinforcing intensity, its dangerous pace pressure, its dual-use diffusion, its uncertainty, and its dangers (pace-versus-safety, escalation, governance challenge, arms-race dynamics) — is essential to understanding AI in national security, because the competition is the context in which military AI is developed and deployed, and it shapes both the technology's trajectory and its risks. The competition is not just a backdrop but an active force creating dangers of its own (especially the pace-versus-safety tension), and analyzing it is central to analyzing military AI's strategic reality. Much of the course's treatment of risks and governance flows from this competitive dynamic, which drives the militarization of AI faster than wisdom about it develops — the gap the module's epigraph names.
Key Insight: The competition among major powers to lead in military AI is a defining strategic dynamic, driven by the perception that AI leadership confers decisive advantage and lagging risks falling dangerously behind, and sustained by its self-reinforcing character (because others compete, none can afford not to). Its dangerous character: pace pressure (to develop and deploy faster than reliability, testing, and governance warrant — trading safety for speed, a central risk), dual-use diffusion (military AI builds on broadly-available commercial AI, easing spread and complicating control), and uncertainty (about how decisive AI really is, driving over-investment and miscalculation). The competition actively creates dangers — pace-versus-safety, escalation, governance challenge, arms-race dynamics — driving militarization faster than wisdom develops.
Military AI raises profound ethical questions, centered on the role of human control and judgment in the use of force, and these questions — which the whole course engages and Module 6 treats in depth — are introduced here as a core dimension of understanding military AI, because they are inseparable from the technology's significance. This section frames the central ethical questions at the analytical level.
The foundational ethical question is the role of human control over the use of force, especially lethal force — how much decision-making about killing should be delegated to machines, and how much must remain with humans. This is the crux of the autonomous-weapons debate (Module 2) and a central issue across military AI: as AI enables systems to operate with more autonomy, including potentially selecting and engaging targets, the question of whether and how much human control and judgment must remain in decisions to use force becomes acute. The positions range from requiring meaningful human control over all use-of-force decisions (the view that lethal decisions must remain human) to accepting greater autonomy in some contexts (the view that autonomous systems may be acceptable or necessary under some conditions) — and this debate about human control over lethal force is perhaps the central ethical question of military AI, engaging deep issues of moral responsibility, human dignity, and the ethics of killing. The course engages this question analytically (Module 2, Module 6), presenting the serious positions and the considerations, as a matter requiring careful ethical analysis rather than a settled answer.
Beyond human control over force, military AI raises further ethical questions the course engages. Accountability: when an AI system is involved in a consequential action (especially a wrongful one — a mistaken killing, an error), who is accountable? — a hard question when the action involved an autonomous or semi-autonomous system, and one central to the ethics and law of military AI. Reliability and the ethics of relying on brittle AI in lethal contexts: given AI's limits (Section 3 — brittleness, lack of judgment, unreliability), the ethics of relying on such AI in life-and-death military decisions is itself a serious question, because AI's failures in lethal contexts are catastrophic and its brittleness in war's novel situations makes failures likely. Compliance with the laws of war: whether and how AI systems can comply with the legal and ethical requirements of armed conflict (distinction between combatants and civilians, proportionality, etc.) — requirements that demand exactly the judgment AI lacks, raising the question of whether AI can meet the legal-ethical standards war's law requires. The lowered threshold for force: whether military AI, by reducing the human cost and effort of using force, lowers the threshold for resorting to force — an ethical and strategic concern that easier, lower-risk force may mean more force.
These ethical questions are inseparable from understanding military AI, and the course engages them as central, not peripheral. Understanding military AI is not just understanding its capabilities and strategic dynamics but engaging the profound ethical questions it raises — human control over force, accountability, the ethics of relying on brittle AI in lethal contexts, compliance with the laws of war, and the threshold for force — because these questions are central to what military AI means and how it should (and should not) be developed and used. The course engages them analytically and evenhandedly (presenting the serious positions and considerations, per the academy's commitment to fair treatment of contested questions), treating them as genuine and difficult ethical questions requiring careful analysis rather than as settled matters — and it treats them as inseparable from the technology, because military AI's ethical dimension is not an add-on but central to its significance. The ethical questions, especially human control over lethal force, are among the most important and contested aspects of military AI, and engaging them clearly and seriously — understanding the questions, the positions, and the considerations — is essential to understanding military AI as the profound and contested development it is. Module 6 develops the governance and ethical frameworks in depth; this section establishes the ethical questions as a core dimension of the whole subject, inseparable from its analysis.
Key Insight: Military AI raises profound ethical questions, central not peripheral to understanding it. The foundational one: the role of human control over the use of force, especially lethal force — how much decision-making about killing may be delegated to machines versus must remain human (the crux of the autonomous-weapons debate, engaging moral responsibility, human dignity, and the ethics of killing). Further questions: accountability (who is responsible when an autonomous system acts wrongly), the ethics of relying on brittle, judgment-lacking AI in lethal contexts, whether AI can comply with the laws of war (distinction, proportionality — requiring the judgment AI lacks), and whether easier AI-enabled force lowers the threshold for using it. The course engages these analytically and evenhandedly, as genuine, difficult, contested questions inseparable from the technology.
This closing section synthesizes the module into a framework for thinking clearly about military AI — the analytical lenses that, applied together, produce sound understanding of any military-AI development or debate, which is the course's aim and the foundation for the modules to come.
Apply these together to any military-AI question:
Applied together, these lenses produce a rounded, honest analysis of any military-AI development — its real capability and limits, its place in the landscape, its strategic dynamics, its risks, its ethical questions, and its governance — which is exactly the clear understanding the course aims to build.
The course's through-line, established here: military AI is a profound, transformative, and contested development — genuinely significant (a general-purpose enhancer across security functions, strategically decisive enough to drive intense competition), genuinely limited (the brittle, judgment-lacking, unreliable AI of the earlier courses, especially risky in war's environment), genuinely dangerous (from its limits meeting war, from the competition's pace pressure, from escalation and control risks), and genuinely contested (raising profound ethical questions, especially about human control over force). Understanding it requires holding all of this honestly — its power and its limits, its strategic significance and its risks, its capabilities and its ethical questions — through the analytical lenses, with the governing commitments to honesty, analysis, evenhandedness, and awareness of the capability-wisdom gap. This framework and these commitments are the foundation for the modules to come, which examine the most significant applications and questions — autonomous weapons, intelligence processing, cyber operations, strategic risks, and governance — each analyzed through these lenses with these commitments, building the clear, honest understanding of AI in national security that the course exists to provide.
Key Insight: Think about military AI through analytical lenses applied together — capability (honest, neither hype nor fear), application (landscape and autonomy level), strategic (competitive dynamics, pace pressure, escalation), risk (from AI's limits meeting war, from competition), ethical (human control over force, accountability, laws of war, threshold), and governance — producing rounded, honest analysis of any development. Hold the governing commitments: honesty over hype and alarm, analytical not operational, evenhandedness on contested questions, and awareness of the capability-wisdom gap (capabilities outrunning our wisdom about them). The through-line: military AI is profound, transformative, and contested — genuinely significant, genuinely limited, genuinely dangerous, genuinely contested — and understanding it requires holding all of this honestly.
You now have the analytical foundation for AI in national security: why AI matters strategically (a general-purpose enhancer, speed-and-scale advantages, the competitive dynamic), the landscape of military AI applications (most of it information-processing and support, with the autonomy dimension distinguishing the contested applications), the honest capability picture (real power, real limits especially dangerous in war), the strategic competition and its dangers (pace-versus-safety, escalation, arms-race dynamics), the central ethical questions (human control over force above all), and the analytical framework and commitments for thinking clearly about it all. You can analyze military AI honestly, neither hyping nor fearing, through the lenses that produce sound understanding.
The next module, Autonomous Weapons Systems, examines the most contested application in depth — the systems that raise the human-control-over-lethal-force question most acutely. You will study, at the analytical and ethical level, what autonomous weapons are and the autonomy spectrum, the serious arguments in the debate over them (the case for and against, presented evenhandedly), the human-control question at its sharpest, and the governance efforts around them. Everything builds on this module's framework: the capability lens (what these systems can and cannot reliably do), the ethical lens (human control, accountability, laws of war), the strategic lens (the competition's pressure), and the governance lens (the efforts to control them) — applied to the application where the stakes and the contest are highest, with the honest, analytical, evenhanded treatment the subject demands.