Who Really Commands? Reflections on Murielle Delaporte’s Series on AI and the Future of Military Command

Murielle Delaporte has been working through one of the most consequential questions facing allied militaries today: what happens to command authority when Artificial Intelligence enters the cognitive loop of the officer making decisions? Her series on AI and military command is not a survey of technology options. It is a sustained analytical inquiry into what it means to lead, and whether the philosophy of Mission Command can survive contact with the machine.

Reading her work in sequence, I am struck by how consistently she returns to a tension that practitioners recognize but that policy documents tend to paper over: the modern battlespace no longer suffers from a shortage of information. It suffers from a paralyzing surplus of it. Every sensor feed, every satellite pass, every tactical terminal adds to what she aptly frames as a “fog of data”, a condition that threatens to overwhelm the decision-making it was designed to illuminate. The commander who once struggled to know enough now struggles to know what matters.

Her central question is whether we can embrace the velocity that AI offers without hollowing out the human soul of leadership. It is the right question, and she pursues it with the rigor of someone who has spent considerable time with the practitioners themselves.

From Tool to Cognitive Partner

One of the more significant contributions in Murielle’s series is her analysis of the shift now underway in how AI is positioned relative to the commander. For years, AI functioned quietly in the background, predictive maintenance, image flagging, logistics optimization. These are valuable capabilities, but they do not alter the cognitive chain of command. What she identifies as genuinely new is the move of AI directly into what she calls the “cognitive loop”: the conversational military assistant that allows an officer to interrogate complex operational data in natural language, without specialized interfaces or coding knowledge.

This matters most, she argues, at the level of decentralized execution. The third pillar of Command and Control has always depended on subordinate leaders having enough situational awareness to act within the commander’s intent without waiting for instruction. AI-assisted query tools lower the barrier for exactly this kind of initiative. A section leader who can interrogate a database as naturally as she would ask a colleague has been handed something real: the capacity to act faster, with more confidence, within the intent of a commander she may not be able to reach. That is a genuine enablement of Mission Command, not a dilution of it if it is used correctly.

The Kill Web Dimension

Murielle’s series dovetails in important ways with work that Ed Timperlake and I have been developing around the concept of the kill web. The shift she describes, from linear kill chains to a dynamic mesh of actors across domains, is precisely the operational architecture that makes the kill web concept necessary. In a service-centric kill chain, decision sequences are manageable because they are essentially linear: find, fix, track, target, engage, assess. In a multi-domain kill web, every action in one domain cascades across the others. Complexity is no longer additive; it is multiplicative.

What Murielle adds to this picture is an account of how AI functions as the essential interface for this mesh, the layer that turns sensor saturation into actionable situational awareness. She traces how an AI-assisted OODA loop operates at machine speed across each phase: automated fusion of multi-source data streams in the Observe phase, rapid cross-referencing of signatures and indicators in the Orient phase, machine-generated course-of-action analysis in the Decide phase, and near-simultaneous synchronization of effects in the Act phase. The kill web without this kind of AI backbone is a concept without operational reality. She is right to press on what it requires.

Training as the Real Proving Ground

Among the most practically useful arguments in her series is her insistence that the real test of AI integration is not combat but training. This is where cognitive habits are formed, where institutional culture is shaped, and where the risks of automation bias either get built in or designed out. She points to adaptive adversary simulations, AI-driven Red Teams that respond dynamically to Blue Force decisions rather than following scripted scenarios, as the kind of training environment that can actually build the cognitive resilience military leaders will need.

Her argument carries a warning embedded in it: if conversational AI is integrated in training as a crutch, if success is consistently rewarded when the officer follows the machine’s recommendation, then Mission Command will wither. The creativity, initiative, and moral courage that distinguish genuine leadership will be optimized away. What she is calling for is AI integration that is demanding, not convenient, tools that force officers to engage critically with machine outputs rather than defer to them. That requires institutional will, and her series is implicitly a call to exercise it.

The Sovereignty Problem

Murielle’s treatment of what she calls the “algorithmic supply chain” is one of the more strategically serious contributions in the series, and it connects directly to debates I have been following among allied defense establishments in Europe and the Indo-Pacific. Sovereignty in the 21st century is no longer primarily about fuel stocks or ammunition reserves. It is about who owns the orchestration platforms, who controls the underlying models, and what political ecosystems those models are embedded in.

The data points she marshals are sobering. The global market share of Chinese AI models has risen sharply in a compressed timeframe, with open-source releases achieving download figures that would have seemed implausible only a few years ago. These models are not neutral instruments. They carry opacity, potential data exfiltration risks, and ties to political systems whose interests diverge from those of allied militaries.

The balance she identifies between the collective effectiveness enabled by allied AI platforms and the national control required for genuine political sovereignty over the decision-making process is one that no allied capital has fully resolved. France’s approach through Artemis.IA is one data point. The NATO Maven Smart System is another.

The tension between them is not incidental; it is structural.

The Human at the Center

What I appreciate most about Murielle’s series is that she refuses the false choice between technophilia and technophobia. She is not arguing that AI must be kept out of the command process, nor that its integration is straightforwardly beneficial. She is arguing for something more demanding: that the doctrinal choices made now will shape whether these tools serve as multipliers for human judgment or substitutes for it.

Her call for AI literacy as a core requirement of Professional Military Education with the same institutional weight as marksmanship or land navigation is practical and necessary. So is her insistence on degraded-environment exercises, on ensuring that officers who are proficient with a Large Language Model interface remain equally proficient with a paper map and compass. The digital backbone will be targeted. The question is whether the human backbone will hold when it is severed.

Murielle is asking who really commands. Her answer, I think, is that this depends entirely on choices that military institutions are making right now, mostly without adequate deliberation. Her series is a contribution to the deliberation that is overdue.