AI Processes Faster. But the Button Needs a Conscience
Wednesday — AI in Defense · DefenseHub
DefenseHub · Wednesday — AI in Defense · June 17, 2026
By R. Planche · Chief Editor & OSINT Curator
Workplace research confirms that relationship-building, gut-level intuition, and moral reasoning remain beyond AI's reach. In defense, those are not soft skills. They are the backbone of alliance cohesion and credible deterrence.
📸 Generated by Leonardo.ia
What We Know
A recent AP News analysis, drawing on workplace experts including Harvard Business School's Marco Iansiti, identifies three categories of human capability that current artificial intelligence cannot replicate: nurturing trust-based relationships, intuitive decision-making rooted in emotional and bodily experience, and moral judgment grounded in genuine conscience. These claims track with broad consensus in AI research that large language models operate through pattern recognition, not embodied experience or moral agency. That consensus is what gives the findings their weight, not any single study, but the convergence of multiple research lines reaching the same conclusion. These three skill gaps map directly onto the most consequential functions in defense. Alliance management depends on trust built over years of personal engagement. Threat assessment at the strategic level relies on commander intuition that integrates ambiguous signals no algorithm can fully parse. And the decision to employ lethal force, especially at the nuclear threshold, demands moral reasoning that no state has been willing to fully delegate to a machine. To answer the blunt question many readers are asking: no, under current U.S. policy, an Awe cannot decide to launch a nuclear weapon. National Command Authority, meaning the President and the Secretary of Defense, retains sole authorization. No autonomous system sits in that chain of command today, and no serious policy proposal would put one there. The reason is not merely technical. It is that nuclear use is the ultimate moral and political act, and delegating it to a machine would strip deterrence of the human judgment that gives it meaning. The distinction matters because the current defense AI conversation is dominated by autonomous targeting, logistics optimization, and sensor fusion. The skills AI lacks are concentrated at the command and diplomatic layers, exactly where mistakes carry existential consequences. That said, the line between "tactical AI" and "strategic human" is not as clean as it sounds. Modern human-machine teaming blurs those layers constantly. An AI system that fuses sensor data and recommends a strike package is already shaping strategic outcomes, even if a human presses the final button. The risk is not that machines replace generals overnight. It is that AI quietly narrows the options commanders see, compressing the space for independent human judgment well before the moment of decision.
Operational Context
This analysis arrives as Congress actively debates who writes the rules for military AI. Legislators are pushing to define the boundaries of autonomous decision-making before the Pentagon locks in its own frameworks. The core tension is familiar: speed versus judgment. AI can process targeting data faster than any human. But the decision to strike, to escalate, to hold fire requires moral and political reasoning that no sensor feed can replace. It is worth noting that not everyone agrees human judgment is automatically more stabilizing. A credible counterargument holds that adversaries might actually find a rational, unemotional AI deterrent more credible, precisely because it would not be subject to panic, fatigue, or domestic political pressure. If an adversary believed an AI system would respond to aggression with cold, optimal precision, that belief could strengthen deterrence rather than weaken it. The problem is that deterrence depends on more than calculated responses. It depends on perceived will, political context, and the adversary's belief that the humans in charge have both the resolve and the restraint to act wisely. A machine that cannot feel the weight of consequences may calculate correctly and still fail to deter, because the adversary sees no conscience behind the threat. Recent patterns in the Indo-Pacific reinforce the point. When tensions between major powers and smaller alliance partners escalate and then subside, de-escalation does not happen by algorithm alone. But AI is not irrelevant to the process. Intelligence systems can surface patterns in military activity, flag anomalies, and provide decision-makers with a clearer operational picture faster than any human staff. That information can buy time and reduce uncertainty, both of which help human leaders choose restraint. What Awe cannot do is manage the political signals that prevent a crisis from tipping over. The phone call between leaders, the quiet diplomatic back-channel, the decision to publicly stand down in a way that gives the other side room to do the same. Those are acts of relational and moral judgment that depend on trust built over years. AI can inform de-escalation. It cannot perform it. The same logic applies to deterrence more broadly. Credible deterrence requires adversaries to believe that the humans behind the weapons possess both the will and the moral framework to act, or not act, under pressure. Delegating that calculus entirely to a system that lacks genuine conscience risks undermining the very psychology that makes deterrence work, even if the system's calculations are technically flawless.
My Read
The question nobody is asking is whether the rush to integrate AI into defense is creating a dangerous illusion of competence at the strategic level. I think it is. When Congress pushes to write rules of engagement for military AI before the Pentagon does, the underlying concern is exactly this. Legislators sense that the technology is outrunning the framework for human accountability. They are right to worry. Alliances are not data structures. NATO cohesion, the Quad, AUKUS, bilateral defense pacts across the Indo-Pacific, all of these rest on personal trust between leaders, diplomats, and military officers built over decades. No AI system replicates that trust. Not this generation. Probably not the next. My bet is that the real risk is not rogue autonomous weapons. It is the slow erosion of human judgment at the command level as organizations lean on AI for speed, gradually sidelining the intuitive and relational skills that actually prevent wars. The danger is subtle: not that a machine overrides a general, but that a generation of officers trained to trust AI recommendations loses the habit of independent strategic thinking. I take the counterargument about rational AI deterrence seriously. If an adversary genuinely believed an AI would respond optimally to every provocation, that might deter certain kinds of aggression. But deterrence is not a math problem. It is a psychological relationship between adversaries, and relationships require the perception of human agency on both sides. If AI demonstrates genuine contextual moral reasoning in high-fidelity wargames, not scripted demos, I would revise. But nothing in the current technology suggests that timeline is short.
What to Watch
Whether the Pentagon's next AI strategy update addresses alliance management and diplomatic decision-support as distinct use cases, or continues to focus almost exclusively on targeting and logistics.
How allied nations, particularly Japan, Australia, and the UK, respond to U.S. AI integration timelines, and whether interoperability concerns slow adoption or force compromises on human-in-the-loop standards.
Whether congressional AI legislation mandates specific human authority thresholds for escalatory decisions, or settles for vague oversight language that leaves discretion to the services.
Any wargame or exercise results that test AI-assisted strategic decision-making against human-only control groups, especially in nuclear or near-peer scenarios.
Results showing AI-informed teams de-escalating more effectively than human-only teams would challenge the core thesis here.
Recommended Sources
RAND Corporation: Extensive research on AI and autonomous systems in defense, with specific work on human-machine teaming and escalation risks in nuclear command and control.
War on the Rocks: Regularly publishes practitioner and academic analysis on the intersection of AI, military ethics, and alliance management at the strategic level.
CNAS (Center for a New American Security): Their AI and autonomy program tracks U.S. and allied policy development on military AI governance, including congressional activity.
RUSI (Royal United Services Institute): Provides a critical allied perspective from the UK on AI integration standards, interoperability challenges, and human-in-the-loop requirements across NATO.
Defense One: Covers congressional defense AI debates and Pentagon acquisition decisions with strong sourcing inside both the legislative and executive branches.
Sources & Methodology
This briefing is based on open-source reporting, official releases, procurement documents, defense-industry disclosures, and specialist analysis available at publication time. Claims involving battlefield effects, classified programs, or active operations are treated cautiously unless corroborated by multiple independent sources.
DefenseHub prioritizes primary sources where available, including official releases, budget and procurement records, legislative documents, technical disclosures, institutional research, and reputable reporting.
Corrections or source clarifications can be sent through the DefenseHub contact page.
— R. Planche · DefenseHub