We may soon face conflict with nation-states in a new model of warfare—one not defined by geography, but instead defined by AIs competing directly with one another. War has always been a ruthless teacher: adaptation, learning, and feedback determine who prevails. In the sixth domain of warfare—AI-driven conflict—that cycle of teaching accelerates beyond human limits. Machine systems do not just execute decisions; they absorb outcomes, update models, and refine strategy continuously and at scale. Strategy, planning, targeting, and execution collapse into a single, ongoing process of learning and acting at speeds no human command structure can match. Missiles, drones, satellites, and fleets remain, but only as instruments—physical endpoints through which AIs prosecute this competition. The real contest is not fought on terrain, but in cycles of learning and adaptation too fast for humans to comprehend. What matters is not force projection, but learning velocity; not command of terrain, but command of inference.
In this domain, doctrines like “human-in-the-loop” or even “human-on-the-loop” are not safeguards—they are handicaps. Any force that inserts human judgment into the learning cycle—whether tactical or strategic—will be systematically outpaced by one that does not. This is not a matter of doctrine or preference, but of competitive logic: systems that act and adapt faster gain an inherent advantage. If removing human constraints yields advantage, those constraints will vanish. This reality forces a stark choice: delegate authority to machines across the full stack of war and accept the risks of opacity and escalation, or retain human control and accept defeat at machine speed. The fog of war has not lifted—it has thickened into a machine-generated battlespace, a black-box system fighting and learning at a level humans can no longer fully observe or comprehend, let alone direct.
This transformation elevates intelligence from a supporting function to the central determinant of power. Models and compute provide capability, but intelligence—continuous, fused inputs from sensors across domains—provides the fuel that drives the system. The intelligence cycle itself describes the change: planning and direction, collection, processing and exploitation, analysis, production, and dissemination. In the sixth domain, machines run all of these functions. In an AI-versus-AI conflict, the side that sees more, sooner, and more clearly directs the learning loop itself. The advantage will not go simply to the best models, but to the systems that integrate intelligence fastest and most effectively. A slightly inferior model, continuously updated with superior, secure, low-latency data, will outperform a more advanced system operating on stale or fragmented inputs. Intelligence no longer informs decisions; it becomes inseparable from them.
The reflexive policy response in the West will be to constrain the leading AI companies in the name of control and safety. That instinct risks slowing the very learning cycles that determine advantage. Winning will require the opposite: accelerating innovation, tightly coupling it to national security objectives, and denying other nation-states key capabilities.
As in the nuclear era, this creates a parallel requirement for stability. The Great Powers share an interest—not in cooperation, but in using the technology to the benefit, rather than detriment, of society. To survive, we all need to understand each other’s AI capabilities and limits. Call it Mutually Assured AI Destruction: MAAID. Like its nuclear predecessor, deterrence will rest not on trust, but on verification, ambiguity, partial visibility, and intelligence where transparency fails. This is mutual self-interest under conditions of instability. In this domain, the equivalent of launch detection—recognizing decisive AI actions at inception—will be critical. Intelligence will need to continuously track, interpret, and outpace adversary AI systems, not only to compete, but to avoid surprise and uncontrolled escalation. What each side can see, and how quickly it can understand it, may determine whether competition remains bounded or spirals beyond control. An AI-era equivalent of the OPCW may therefore become necessary—an institution capable of inspecting, verifying, and monitoring advanced weapons AI systems under conditions of speed and opacity.
In the near future, humans will at best set initial conditions—policy, objectives, constraints—and then watch as their systems prosecute a war they can no longer meaningfully steer. Clausewitz wrote that war is the continuation of policy by other means. In the sixth domain, that distinction collapses. When systems translate intent into action, learn from the results, and adjust course continuously at machine speed, policy is no longer something periodically expressed through war. It is instantiated in code and executed as a continuous process. The side that learns fastest does not just win—it shapes the war itself.
The Cipher Brief is committed to publishing a range of perspectives on national security issues submitted by deeply experienced national security professionals. Opinions expressed are those of the author and do not represent the views or opinions of The Cipher Brief.
Have a perspective to share based on your experience in the national security field? Send it to Editor@thecipherbrief.com for publication consideration.
Read more expert-driven national security insights, perspective and analysis in The Cipher Brief
