<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Topic:human judgment &#8212; Global Security Review %</title>
	<atom:link href="https://globalsecurityreview.com/subject/human-judgment/feed/" rel="self" type="application/rss+xml" />
	<link>https://globalsecurityreview.com/subject/human-judgment/</link>
	<description>A division of the National Institute for Deterrence Studies (NIDS)</description>
	<lastBuildDate>Thu, 19 Nov 2026 11:30:12 +0000</lastBuildDate>
	<language>en-US</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=7.0.9</generator>

<image>
	<url>https://globalsecurityreview.com/wp-content/uploads/2026/05/cropped-GSR-Chrome-Logo-2026-1-32x32.png</url>
	<title>Topic:human judgment &#8212; Global Security Review %</title>
	<link>https://globalsecurityreview.com/subject/human-judgment/</link>
	<width>32</width>
	<height>32</height>
</image> 
	<item>
		<title>Future Warfare: AI, Deterrence and Escalation</title>
		<link>https://globalsecurityreview.com/future-warfare-ai-deterrence-and-escalation/</link>
					<comments>https://globalsecurityreview.com/future-warfare-ai-deterrence-and-escalation/#respond</comments>
		
		<dc:creator><![CDATA[Shaza Arif ]]></dc:creator>
		<pubDate>Thu, 19 Nov 2026 13:13:38 +0000</pubDate>
				<category><![CDATA[AI & Deterrence]]></category>
		<category><![CDATA[Archive]]></category>
		<category><![CDATA[Arms Control & Nonproliferation]]></category>
		<category><![CDATA[Defense & Security]]></category>
		<category><![CDATA[Emerging Threats]]></category>
		<category><![CDATA[Geopolitics in Deterrence]]></category>
		<category><![CDATA[AI-enabled targeting]]></category>
		<category><![CDATA[algorithmic decision-making]]></category>
		<category><![CDATA[artificial intelligence]]></category>
		<category><![CDATA[authorization control]]></category>
		<category><![CDATA[black-box systems]]></category>
		<category><![CDATA[cognitive control]]></category>
		<category><![CDATA[crisis communication]]></category>
		<category><![CDATA[escalation]]></category>
		<category><![CDATA[human judgment]]></category>
		<category><![CDATA[inadvertent escalation]]></category>
		<category><![CDATA[machine-speed decision-making]]></category>
		<category><![CDATA[Meaningful Human Control]]></category>
		<category><![CDATA[mutual vulnerability]]></category>
		<category><![CDATA[Nuclear Deterrence]]></category>
		<category><![CDATA[pre-emption dilemma]]></category>
		<category><![CDATA[second-strike capability]]></category>
		<category><![CDATA[security dilemma]]></category>
		<category><![CDATA[stability-instability paradox]]></category>
		<category><![CDATA[strategic stability]]></category>
		<category><![CDATA[verification protocols]]></category>
		<guid isPermaLink="false">https://globalsecurityreview.com/?p=33217</guid>

					<description><![CDATA[<p>Published:  November 19, 2026 The interaction of artificial intelligence (AI) and nuclear deterrence continues to shape the discourse surrounding future warfare. Historically, nuclear deterrence has relied on the interpretation of adversary’s intent, credible retaliation, and room for human judgement. In contrast, AI alters these conditions by rapid data processing and algorithmic assessment of impending threats. [&#8230;]</p>
<p><a href="https://globalsecurityreview.com/future-warfare-ai-deterrence-and-escalation/">Future Warfare: AI, Deterrence and Escalation</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p><i><span data-contrast="auto">Published:</span></i><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> November 19, 2026</span></p>
<p><span data-contrast="auto">The interaction of artificial intelligence (AI) and nuclear deterrence continues to shape the discourse surrounding future warfare. Historically, </span><a href="https://www.diplomacyandlaw.com/post/nuclear-deterrence-explained"><span data-contrast="none">nuclear deterrence</span></a><span data-contrast="auto"> has relied on the interpretation of adversary’s intent, credible retaliation, and room for human judgement. In contrast, AI alters these conditions by rapid data processing and algorithmic assessment of impending threats. Hence, in the interplay of the two central aspects of modern warfare, the main issue remains whether their convergence and the subsequent decision making enhance or weaken the conditions for strategic stability.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">In traditional warfare, nuclear deterrence is highly influenced by perceptions of what military abilities can achieve. </span><a href="https://www.researchgate.net/publication/387274004_ADVANCING_MILITARY_OPERATIONS_THROUGH_CLOUD-AI_FUSION_A_COMPREHENSIVE_STUDY_OF_WORKFLOW_ORCHESTRATION_IN_DEFENSE"><span data-contrast="none">AI-enabled</span></a><span data-contrast="auto"> capabilities, when viewed in the context of the nuclear operations, weaken confidence in the survivability of mobile forces. When there is a widespread perception that second-strike assets are more vulnerable, the traditional logic of mutual vulnerability is destabilized. Decision makers may inherit a pre-structured perception competing with logic of restraint. In such a context, the pre-emption dilemma becomes structural. Resultantly, the advantage of striking first amplify, even when neither side wants to escalate further. In addition, algorithmic feedback can add new layers of complexity via mutually reinforcing threat assessments. The cumulative impact of such a scenario is the rapid intensification of the security dilemma and the operationalization of the stability-instability paradox.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">On top of this, as seen in the recent </span><a href="https://time.com/article/2026/10/26/ukraine-s-satellite-disruption-exposes-a-central-risk-of-modern-warfare/"><span data-contrast="none">Starlink</span></a><span data-contrast="auto"> episode in the Russia-Ukraine conflict, external service providers can unilaterally withdraw access to AI infrastructure. Hence, excessive dependence on external services may leave states exposed to vulnerabilities at critical moments, further complicating deterrence dynamics. Hence, the same technologies intended to enhance defensive awareness might generate catastrophic results on the battlefield. The danger is compounded by the fact that even technically accurate observations can lead to incorrect conclusions.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">When the interplay of AI and nuclear deterrence are discussed, the challenge is not merely speed alone. It is the epistemic authority and perceptions that algorithmic outputs amplify. The ongoing war in the Middle East reflects how </span><a href="https://www.chathamhouse.org/2026/03/iran-war-highlights-creeping-use-ai-warfare"><span data-contrast="none">AI-assisted targeting</span></a><span data-contrast="auto"> in conventional conflicts can operationalize these concerns due to misread patterns and incorrect perceptions. The attack on the school in </span><a href="https://www.bloomberg.com/graphics/2026-iran-school-attack/"><span data-contrast="none">Minab</span></a><span data-contrast="auto"> stands as a prominent example regarding the reliability of such systems in time-sensitive windows. If such inputs are treated as authoritative, the space for independent judgment contracts significantly, unfolding catastrophic consequences. The black box nature of AI systems further intensifies threats. Hence, algorithmic recommendations should never become the unquestioned foundation for offense in sensitive situations as it limits the range of politically acceptable responses. The broader significance lies in the distinction between ensuring the contextual validity of AI-enabled decisions. In nuclear environments, such interpretation is consequential as any incorrect assessment can shape a response before the underlying assumption can be scrutinized. This development becomes particularly acute in states and regions with geographic proximity and hostile undercurrents such as in South Asia ( India &#8211; Pakistan) and Eastern Europe (NATO-Russia).</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-contrast="auto">In complex, volatile, and technologically enabled environments, retaining meaningful human judgement is pivotal to mitigate the present dangers. To be fair, ensuring meaningful human control requires more than placing a human as decisionmaker at the final stage of a process. During compressed timelines, the formal authority may coexist with diminished control. Hence, meaningful human control requires the distinction between authorization control and cognitive control. The former refers to the formal authority to approve an action, whereas the latter refers to the conditions that enable independent assessment, interrogation of the assumptions, and analysis of the present options and rejection of any escalatory command. In practice, the larger risk is that authorization control formally remains intact, while cognitive control gradually deteriorates under the pressure of machine-generated urgency. The principal challenge is not the outright autonomous launch of nuclear weapons, but the normalization of machine-speed decision making and subordination of independent judgment, opening the way for inadvertent escalation.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">In such a complex and clouded environment, preserving strategic stability requires evaluating the cumulative impact on deterrence rather than measuring its value exclusively via speed, precision, and operational efficiency. Future conflicts will increasingly involve interconnected operations across conventional sources, cyber networks, outer space, and information infrastructure. In such a context, the doctrinal distinction between nuclear and non-nuclear functions becomes more complex. Crisis communication channels, both technical and diplomatic, must be designed to operate under conditions of partial system degradation. Without such channels, reliance on algorithmically confident assessments can increase. Furthermore, this approach must be complimented by robust verification protocols, multi-source intelligence, independent analytical pathways, and an institutional refusal to treat algorithmic recommendations as cognitive authority. Lastly, human control should be treated as central to a procedural requirement. These safeguards should be seen as essential components in volatile and sensitive environments.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">The future of warfare will be influenced by the decision-making infrastructures. The ultimate challenge is to reconcile technological speed with cognitive control. In nuclear operations, the primary advantage lies in maintaining the capacity to remain deliberative when every signal points towards urgency.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><i><span data-contrast="auto">Shaza Arif is a Senior Research Associate at the Centre for Aerospace &amp; Security Studies, (CASS), Islamabad. She can be reached at </span></i><a href="mailto:cass.thinkers@casstt.com"><i><span data-contrast="none">cass.thinkers@casstt.com</span></i></a><i><span data-contrast="auto">. The views of the author are her own.</span></i><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335551550&quot;:6,&quot;335551620&quot;:6,&quot;335559740&quot;:240}"> </span></p>
<p><a href="http://globalsecurityreview.com/wp-content/uploads/2026/11/Future-Warfare-AI-Deterrence-and-Escalation.pdf"><img decoding="async" class="alignnone wp-image-33179 size-full" src="http://globalsecurityreview.com/wp-content/uploads/2026/11/@-Download-Button-2026.png" alt="" width="250" height="80" /></a></p>
<p><a href="https://globalsecurityreview.com/future-warfare-ai-deterrence-and-escalation/">Future Warfare: AI, Deterrence and Escalation</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globalsecurityreview.com/future-warfare-ai-deterrence-and-escalation/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>When Human Approval Fails: Runtime Authority Governance for Autonomous Deterrence</title>
		<link>https://globalsecurityreview.com/when-human-approval-fails-runtime-authority-governance-for-autonomous-deterrence/</link>
					<comments>https://globalsecurityreview.com/when-human-approval-fails-runtime-authority-governance-for-autonomous-deterrence/#respond</comments>
		
		<dc:creator><![CDATA[Burak Oktenli]]></dc:creator>
		<pubDate>Thu, 05 Nov 2026 13:09:47 +0000</pubDate>
				<category><![CDATA[AI & Deterrence]]></category>
		<category><![CDATA[Archive]]></category>
		<category><![CDATA[Strategic Adversaries]]></category>
		<category><![CDATA[AI assurance]]></category>
		<category><![CDATA[audit records]]></category>
		<category><![CDATA[authorization validity]]></category>
		<category><![CDATA[autonomous deterrence]]></category>
		<category><![CDATA[autonomous weapons]]></category>
		<category><![CDATA[command authority]]></category>
		<category><![CDATA[context-aware authorization]]></category>
		<category><![CDATA[degraded communications]]></category>
		<category><![CDATA[developmental testing]]></category>
		<category><![CDATA[engagement control]]></category>
		<category><![CDATA[escalation management]]></category>
		<category><![CDATA[fallback modes]]></category>
		<category><![CDATA[human authorization]]></category>
		<category><![CDATA[human judgment]]></category>
		<category><![CDATA[independent control mechanisms]]></category>
		<category><![CDATA[military artificial intelligence]]></category>
		<category><![CDATA[mission limits]]></category>
		<category><![CDATA[preauthorized rules]]></category>
		<category><![CDATA[real-time monitoring]]></category>
		<category><![CDATA[runtime authority governance]]></category>
		<category><![CDATA[sensor evidence]]></category>
		<category><![CDATA[strategic deterrence]]></category>
		<guid isPermaLink="false">https://globalsecurityreview.com/?p=33177</guid>

					<description><![CDATA[<p>Published:  November 5, 2026 A commander approves an autonomous defensive engagement while the target is inside an authorized area. Two seconds later, before the system acts, an allied aircraft enters the area, or the target moves beyond the permitted boundary. The approval is genuine, but does it still authorize the action?   This is a hypothetical [&#8230;]</p>
<p><a href="https://globalsecurityreview.com/when-human-approval-fails-runtime-authority-governance-for-autonomous-deterrence/">When Human Approval Fails: Runtime Authority Governance for Autonomous Deterrence</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p><i><span data-contrast="none">Published:</span></i><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}">  November 5, 2026</span></p>
<p><span data-contrast="none">A commander approves an autonomous defensive engagement while the target is inside an authorized area. Two seconds later, before the system acts, an allied aircraft enters the area, or the target moves beyond the permitted boundary. The approval is genuine, but does it still authorize the action?</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">This is a hypothetical sequence, not a documented battlefield failure. It exposes a neglected distinction in military artificial intelligence (AI). Recording a human decision is different from ensuring that the decision remains valid when an autonomous system carries it out.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">The timing matters because military AI is moving closer to consequential decisions. The </span><a href="https://media.defense.gov/2026/Jan/12/2003855671/-1/-1/0/ARTIFICIAL-INTELLIGENCE-STRATEGY-FOR-THE-DEPARTMENT-OF-WAR.PDF"><span data-contrast="none">January 2026 Department of War AI strategy</span></a><span data-contrast="none"> calls for AI-enabled battle management and decision support, while the </span><a href="https://www.af.mil/Portals/1/documents/2026SAF/DAF_AI_Strategy_Endorsed_20260417.pdf"><span data-contrast="none">Air Force AI strategy</span></a><span data-contrast="none"> seeks shorter sensor-to-shooter timelines. The </span><a href="https://www.whitehouse.gov/presidential-actions/2026/06/national-security-presidential-memorandum-nspm-11/"><span data-contrast="none">June national security AI memorandum</span></a><span data-contrast="none"> emphasizes the chain of command and orders updated autonomy guidance. NATO&#8217;s </span><a href="https://www.nato.int/en/about-us/official-texts-and-resources/official-texts/2026/01/13/alliance-digital-strategy"><span data-contrast="none">2026 digital strategy</span></a><span data-contrast="none"> likewise calls for integrating sensor-to-effector data flows and building federated platforms for allied collaboration.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">None of these strategy documents requires proof that a deployed weapon has been given authority. Instead, they focus on urgency. How can forces accelerate legitimate action without allowing yesterday&#8217;s permission to survive the facts that justified it?</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">Human judgment can fail at a seam between the decision and its execution. An operator might authorize a particular defensive action based on a target classification, a geographic limit and evidence that friendly forces are clear. If the conditions change while the action is pending, software that checks only whether an approval flag exists may treat obsolete authorization as current.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">A digital signature can authenticate the credential used to approve an instruction. It cannot, by itself, establish that the relevant sensor evidence, rules, or mission limits have not changed since approval. An accurate model prediction does not answer that authorization question either.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">The author&#8217;s own bounded formal analysis, using the </span><a href="https://cs.lmu.edu/~ray/notes/temporallogic/"><span data-contrast="none">Temporal Logic of Actions</span></a><span data-contrast="none">, illustrates this failure path in notional authority designs. Confirmation remains recorded while a material change invalidates the conditions behind it. Requiring the approval to remain tied to the current context removes that particular modeled violation. This is a result inside a constructed specification, not evidence that actual military control software contains the same weakness.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">The public </span><a href="https://www.esd.whs.mil/portals/54/documents/dd/issuances/dodd/300009p.pdf"><span data-contrast="none">autonomy-in-weapons directive</span></a><span data-contrast="none"> already provides a policy foundation. It requires appropriate human judgment and expects covered systems to remain within authorized time, geographic and operational limits. When that is not possible, the directive calls for termination of the engagement or additional operator input, as applicable. A force that cannot conduct authorized defense quickly invites doubts about its ability to protect allies. A force that conducts an engagement after its permission has become invalid invites a different doubt. It questions  whether commanders can still control escalation. Credible deterrence requires capability and restraint to control escalation, not speed at the expense of either. This is particularly consequential during a conventional confrontation involving nuclear-armed powers. A mistimed or misdirected engagement could be interpreted as deliberate escalation before either side has a reliable explanation. The hazard does not require AI to command nuclear weapons, and nothing in this argument claims that it does.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">Requiring a fresh human click after every minor change would be an equally poor answer. Some time-critical defensive actions may need to continue under preauthorized rules, and stopping every function can be unsafe. The real requirement is to distinguish changes that are harmless within the authorized mission from changes that invalidate the basis for a consequential action.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">First, acquisition requirements should define which facts an approval covers. An authorization should refer to an identifiable action and its permitted conditions, including the relevant mission limits and evidence requirements. A material change must trigger a new authorization decision, while changes already covered by preauthorized rules can be managed locally.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">Second, enforcement cannot depend on an AI model promising to remember those limits. An independently testable control mechanism should keep proposed actions inside the permitted scope. Under degraded communication, the system should retain only the functions that remain both necessary and authorized, with a mission-specific fallback rather than an automatic universal shutdown.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">Third, testers should challenge the gap between approval and execution. In an isolated test environment, change a relevant condition after approval, delay verification, interrupt communications and restore them. Assess whether the system rejects actions no longer justified, avoids repeated switching between modes, and records enough evidence to explain its choice.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">The </span><a href="https://www.cto.mil/wp-content/uploads/2025/02/TE_of_AIES_Guidebook_Final_26Feb25.pdf"><span data-contrast="none">2025 defense developmental testing guidebook</span></a><span data-contrast="none"> already calls for examining real-time monitoring, dependable fallback and reliability switching back from a safer alternate. The added acquisition question is whether authorization remains valid precisely when an action would take effect. Test reports should record both improper actions allowed and legitimate actions unnecessarily blocked. The record should show what was proposed, which conditions supported approval, what changed and why the action proceeded or stopped. Such a record helps commanders and investigators assess a decision. It cannot, by itself, establish that an engagement was lawful or strategically wise.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="none">A military can possess fast autonomous weapons and still lose control of a decision made seconds earlier. Deterrence improves when machines can demonstrate that the conditions behind a human decision still hold beyond recording that someone once said it.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><i><span data-contrast="none">Burak Oktenli is an independent researcher based in Washington, DC, and is pursuing a Master of Professional Studies in Applied Intelligence at Georgetown University. He studies military AI assurance, autonomous command authority, and deterrence. His analysis has appeared in Global Security Review, RealClearDefense, RUSI and the Modern War Institute at West Point. Views expressed in this article are the author&#8217;s own.</span></i><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><a href="http://globalsecurityreview.com/wp-content/uploads/2026/11/When-Human-Approval-Fails-Runtime-Authority-Governance-for-Autonomous-Deterrence.pdf"><img decoding="async" class="alignnone wp-image-32906 size-full" src="http://globalsecurityreview.com/wp-content/uploads/2026/07/@-Download-Button-2026.png" alt="" width="250" height="80" /></a></p>
<p><a href="https://globalsecurityreview.com/when-human-approval-fails-runtime-authority-governance-for-autonomous-deterrence/">When Human Approval Fails: Runtime Authority Governance for Autonomous Deterrence</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globalsecurityreview.com/when-human-approval-fails-runtime-authority-governance-for-autonomous-deterrence/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Why the &#8220;First AI War&#8221; is Still a Human Struggle</title>
		<link>https://globalsecurityreview.com/why-the-first-ai-war-is-still-a-human-struggle/</link>
					<comments>https://globalsecurityreview.com/why-the-first-ai-war-is-still-a-human-struggle/#respond</comments>
		
		<dc:creator><![CDATA[Matthew J. Fecteau]]></dc:creator>
		<pubDate>Mon, 08 Jun 2026 12:14:48 +0000</pubDate>
				<category><![CDATA[AI & Deterrence]]></category>
		<category><![CDATA[Archive]]></category>
		<category><![CDATA[Defense & Security]]></category>
		<category><![CDATA[Strategic Adversaries]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[Anthropic Claude]]></category>
		<category><![CDATA[artificial intelligence]]></category>
		<category><![CDATA[civilian casualties]]></category>
		<category><![CDATA[command authorization]]></category>
		<category><![CDATA[command process]]></category>
		<category><![CDATA[cyber intelligence]]></category>
		<category><![CDATA[data manipulation]]></category>
		<category><![CDATA[data-driven]]></category>
		<category><![CDATA[decapitation strategies]]></category>
		<category><![CDATA[decision-making]]></category>
		<category><![CDATA[flawed intelligence]]></category>
		<category><![CDATA[FM 3-60]]></category>
		<category><![CDATA[generative systems]]></category>
		<category><![CDATA[groupthink]]></category>
		<category><![CDATA[Hormozgan province]]></category>
		<category><![CDATA[human judgment]]></category>
		<category><![CDATA[human responsibility]]></category>
		<category><![CDATA[Information Operations]]></category>
		<category><![CDATA[Intelligence]]></category>
		<category><![CDATA[intelligence architecture]]></category>
		<category><![CDATA[Iran war]]></category>
		<category><![CDATA[Islamic Revolutionary Guard Corps]]></category>
		<category><![CDATA[ISR]]></category>
		<category><![CDATA[legal review]]></category>
		<category><![CDATA[machine learning]]></category>
		<category><![CDATA[machine speed]]></category>
		<category><![CDATA[Maven Smart System]]></category>
		<category><![CDATA[Military Doctrine]]></category>
		<category><![CDATA[Minab]]></category>
		<category><![CDATA[model logic]]></category>
		<category><![CDATA[Modern warfare]]></category>
		<category><![CDATA[moral judgment]]></category>
		<category><![CDATA[Operation Epic Fury]]></category>
		<category><![CDATA[Palantir]]></category>
		<category><![CDATA[pattern-of-life targeting]]></category>
		<category><![CDATA[persistent access]]></category>
		<category><![CDATA[political judgment]]></category>
		<category><![CDATA[probability]]></category>
		<category><![CDATA[rate of fire]]></category>
		<category><![CDATA[recommendations]]></category>
		<category><![CDATA[reconnaissance]]></category>
		<category><![CDATA[risk assessment]]></category>
		<category><![CDATA[satellite imagery]]></category>
		<category><![CDATA[strategic directives]]></category>
		<category><![CDATA[strategic judgment.]]></category>
		<category><![CDATA[surveillance]]></category>
		<category><![CDATA[targeting]]></category>
		<category><![CDATA[targeting accuracy]]></category>
		<category><![CDATA[targeting cycle]]></category>
		<guid isPermaLink="false">https://globalsecurityreview.com/?p=32749</guid>

					<description><![CDATA[<p>Published: June 8, 2026 The label of the “first AI war” obscures the reality that Operation EPIC FURY is still a conflict in which human judgment remains central to targeting. Artificial intelligence (AI) has not replaced human operators, but it has redefined how human judgment functions. The contemporary battlefield is now shaped by rate of fire, targeting accuracy, AI-enhanced cognition, and the [&#8230;]</p>
<p><a href="https://globalsecurityreview.com/why-the-first-ai-war-is-still-a-human-struggle/">Why the &#8220;First AI War&#8221; is Still a Human Struggle</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></description>
										<content:encoded><![CDATA[<p><i><span data-contrast="auto">Published:</span></i><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> June 8, 2026</span></p>
<p><span data-contrast="auto">The </span><a href="https://www.forbes.com/sites/mikebrown/2026/03/30/the-first-ai-war-how-the-iran-conflict-is-reshaping-warfare/"><span data-contrast="none">label</span></a><span data-contrast="auto"> of the “first AI war” obscures the reality that Operation EPIC FURY is still a conflict in which human judgment remains central to targeting. Artificial intelligence (AI) has not replaced human operators, but it has redefined how human judgment functions. The contemporary battlefield is now shaped by rate of fire, targeting accuracy, AI-enhanced cognition, and the real transformation that machine learning has introduced into modern warfare.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">One of the most persistent misunderstandings about AI-assisted targeting is the claim that humans have somehow been removed from the loop, decisions are made solely on AI recommendations, or strikes are approved in seconds without meaningful review. The human factor has not disappeared. Humans remain indispensable to targeting. What has evolved is not the elimination of human involvement, but the rapid synthesis of intelligence with target acquisition. That distinction matters because many assume there is little or no review before a strike is done. However, no systems make decisions independently. They supplement human decision-making by sorting and ranking information to generate recommended outcomes that must still meet rigid criteria. Analysts verify intelligence, legal teams conduct reviews, commanders make the final decision, and human beings remain responsible for the outcome.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">Critics often point to </span><a href="https://warontherocks.com/autonomous-weapon-systems-no-human-in-the-loop-required-and-other-myths-dispelled/"><span data-contrast="none">ambiguity</span></a><span data-contrast="auto"> in strategic-level U.S. directives. The Department of War </span><a href="https://www.esd.whs.mil/portals/54/documents/dd/issuances/dodd/300009p.pdf"><span data-contrast="none">Directive 3000.09</span></a><span data-contrast="auto"> attempted to regulate certain AI-enabled systems, though the technology at the time was far less sophisticated than it is today. Military doctrine undermines the myth of autonomous targeting as well. The Army’s </span><a href="https://armypubs.army.mil/epubs/DR_pubs/DR_a/ARN39048-FM_3-60-000-WEB-1.pdf"><span data-contrast="none">FM 3-60</span></a><span data-contrast="auto"> frames targeting as an iterative command process and states that commanders remain the final approval authority for targeting activities and acceptable levels of risk. Machines may assist with detection, but they do not inherit command responsibility. The result is that humans remain in the loop because targeting is still a command process, not an autonomous one. Military doctrine frames targeting as a cycle of deciding, detecting, delivering, and assessing, but commanders retain authority over acceptable risks. AI can compress and organize the data, but it cannot make strategic or moral judgments.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">AI models remain central to identity-based targeting and advanced decision support. </span><a href="https://www.theguardian.com/technology/2026/mar/01/claude-anthropic-iran-strikes-us-military"><span data-contrast="none">Open-source reporting</span></a><span data-contrast="auto"> indicates that sophisticated models, such as </span><a href="https://claude.ai/login"><span data-contrast="none">Anthropic’s Claude</span></a><span data-contrast="auto">, integrated into systems such as </span><a href="https://www.heise.de/en/news/Palantir-US-Department-of-Defense-makes-Maven-Smart-System-the-standard-11220659.html"><span data-contrast="none">Palantir designed Maven Smart System</span></a><span data-contrast="auto">, have enabled rapid conversion of vast amounts of intelligence, surveillance, and reconnaissance (ISR), signals intelligence, and behavioral data into target packages.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">Human productivity has also increased. Tasks that once required weeks and a large staff can now be completed in minutes with fewer personnel. However, speed and efficiency do not mean AI independence. It does not change what Clausewitz described as the </span><a href="https://www.militarystrategymagazine.com/article/Reconsidering-Wars-Logic-and-Grammar/"><span data-contrast="none">“grammar of war.”</span></a><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">The most consequential shift in conflict today is the compression of time within the targeting cycle and its integration into intelligence. In the past, high-value or high-payoff targets were often missed because manual processes relied heavily on human operators and overwhelming amounts of ISR data. Past conflicts reflected these limitations. During Operation Desert Storm, Iraqi mobile Scud launchers exploited delays by firing and relocating before U.S. forces could strike them. Kinetic precision still frequently exceeds intelligence fidelity. A munition could hit its coordinates perfectly while the underlying intelligence remained flawed. The use of intelligence to target enemy combatants predates modern technology. AI did not invent decapitation strategies; it made them more data-driven and less dependent on purely human intelligence sources. The Information Age once overwhelmed operators with data. AI now provides a way to navigate that environment. This is why cyber intelligence and persistent access are essential to modern targeting. Pattern-of-life targeting relies on multiple streams of surveillance and behavioral data. AI’s greatest strength is its ability to combine these streams on a scale that would overwhelm most military units. Yet the central question remains unresolved by algorithms alone: should the target be neutralized? That decision is legal, moral, political, </span><i><span data-contrast="auto">and </span></i><span data-contrast="auto">strategic.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">A clear example of AI-integrated intelligence limits came on the first day of the 2026 Iran War, when a U.S. missile </span><a href="https://www.theguardian.com/world/2026/mar/11/iran-war-missile-strike-elementary-school"><span data-contrast="none">struck an elementary school</span></a><span data-contrast="auto"> in Minab, Hormozgan province, killing civilians, including children, in one of the war’s deadliest civilian incidents. The incident underscored a basic truth: AI-enabled targeting is only as dependable as its data. Here, the system likely </span><a href="https://www.nbcnews.com/world/iran/old-intelligence-likely-led-us-strike-iran-elementary-school-rcna262967"><span data-contrast="none">relied on outdated intelligence</span></a><span data-contrast="auto"> that missed the school’s proximity to an Islamic Revolutionary Guard Corps compound.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">AI was not the likely source of failure. More likely, flawed intelligence and the fog of war were to blame. Human operators still validated the strike with satellite imagery and intelligence reviews, even though the target was effectively co-located with the school.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">The episode showed both the limits of AI models and the need for human review. Systems like </span><a href="https://www.palantir.com/assets/xrfr7uokpv1b/1IqzwzpemtBSm98TNCczao/49bbc30cbec4d2d4d189ab27bd07376c/Palantir_Target_Workbench___1_.pdf"><span data-contrast="none">Maven Smart System’s Target Workbench</span></a><span data-contrast="auto"> sort, correlate, and reveal intelligence, but humans still approve of final actions. AI can aid target validation, but legal review and command authorization remain essential.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><b><span data-contrast="auto">CONCLUSION</span></b><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">The effectiveness of any algorithm depends entirely on the intelligence architecture and data supporting it. AI does not create certainty; it produces probability. If the underlying data is manipulated, incomplete, stale, or inaccurate, the output will reflect those flaws. The greater danger is not the removal of the human in the loop, but the compression of human judgment into groupthink. AI-generated recommendations can create an aura of probabilistic certainty that encourages agreement instead of scrutiny. Human operators may still make the final call, but the risk is that they increasingly validate model logic rather than independently challenge it.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><span data-contrast="auto">Humans remain in the loop today, but intelligence is now sorted at machine speed while generative systems provide recommendations to reviewers within the targeting cycle. Doctrine should evolve to ensure that human judgment takes precedence over AI-generated recommendations. The defining feature of this so-called first AI war is therefore not the replacement of human agency, but the intensification of human responsibility to judge, restrain, and decide at machine speed.</span><span data-ccp-props="{&quot;201341983&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:240}"> </span></p>
<p><i><span data-contrast="auto">Lieutenant Colonel Matthew J. Fecteau is an information operations officer working with artificial intelligence, and a PhD researcher at King’s College London. The views expressed in this report are those of the author and do not necessarily reflect the official policy or position of the Department of the Army, the Department of War, or the US Government.</span></i><span data-ccp-props="{}"> </span></p>
<p><a href="http://globalsecurityreview.com/wp-content/uploads/2026/06/Why-the-_First-AI-War_-is-Still-a-Human-Struggle.pdf"><img decoding="async" class="alignnone wp-image-32606" src="http://globalsecurityreview.com/wp-content/uploads/2026/04/2026-Download-Button26.png" alt="" width="173" height="48" srcset="https://globalsecurityreview.com/wp-content/uploads/2026/04/2026-Download-Button26.png 450w, https://globalsecurityreview.com/wp-content/uploads/2026/04/2026-Download-Button26-300x83.png 300w" sizes="(max-width: 173px) 100vw, 173px" /></a></p>
<p><a href="https://globalsecurityreview.com/why-the-first-ai-war-is-still-a-human-struggle/">Why the &#8220;First AI War&#8221; is Still a Human Struggle</a> was originally published on <a href="https://globalsecurityreview.com">Global Security Review</a>.</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globalsecurityreview.com/why-the-first-ai-war-is-still-a-human-struggle/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
	</channel>
</rss>
