On August 26, President Donald Trump declared a national emergency over foreign actors exploiting vulnerabilities in the U.S. bulk-power system. The order framed the problem as one of national security, public safety, and economic resilience. A month later, CISA issued new guidance on cyber decoys to help critical-infrastructure defenders catch intruders using legitimate credentials and native tools. Those moves improve prevention and detection. They do not answer the harder deterrence question: what happens after defenders discover that a sophisticated adversary is already inside? 

If removing hostile access would also halt power, logistics, communications, or another essential service, the intrusion can create leverage before a destructive command is ever issued. The strategic problem is therefore larger than penetration. It is whether the defender can withdraw trust without inflicting the disruption the adversary hoped to create. 

 A Foothold Can Coerce Without Firing 

The U.S. has already documented the pre-positioning problem. A joint federal fact sheet on Volt Typhoon says the People’s Republic of China-linked group compromised organizations in communications, energy, transportation, and water systems to position itself for potential disruption during a future crisis. Pre-positioning matters because an unused foothold can still shape decisions if leaders believe evicting it will damage their own mission.  

This is a form of coercive dependence. A defender can become so reliant on a remote connection, cloud service, supplier, identity system, or shared data source that severing the compromised path appears nearly as costly as leaving it in place. An adversary then gains value from access even without activating a destructive effect.  

Selective Disconnection Is Deterrence by Denial  

Joint CI Fortify guidance released in July addresses the operational side of this problem. It urges critical-infrastructure operators to prepare to isolate vital operational technology and enabling systems while maintaining essential operations. Britain’s operational-technology guidance likewise distinguishes full site isolation from service-specific isolation and hardware-enforced trusted communications.  

The strategic value is in the word selective. A suspected component may still be allowed to observe while losing permission to issue commands, change configurations, move resources, or communicate outside a defined boundary. The goal is to remove the adversary’s options faster than the defender removes its own mission.  

This is deterrence by denial at the level of system architecture. An attacker that expects access to produce a reliable ability to stop the mission gains less from the intrusion if the defender can rapidly reduce that access to a narrow, monitored state while essential functions continue.  

The Disconnect Control Can Be Attacked Too  

Isolation is not automatically safe. False alarms, manipulated telemetry, or ambiguous indicators can push defenders toward unnecessary shutdowns and hand an adversary disruption without a successful destructive action. Exercises should therefore evaluate both genuine compromise and misleading evidence, including who can impose restrictions, which facts justify the decision, and how an incorrect restriction is reversed.  

The same discipline applies to the control plane itself. The authority to isolate a site, revoke a supplier, or cut a management path should not depend on the same compromised credential or platform it is meant to constrain. Separation of duties and independent control paths turn disconnection from a policy promise into an operational capability.  

Recovery Must Be Earned  

Disconnection is only half of the problem. A defender that can isolate a compromised dependency but cannot establish when reconnection is justified may trade adversary leverage for self-imposed paralysis. NIST’s cyber-resiliency engineering guidance defines resilience around the ability to anticipate, withstand, recover from, and adapt to attacks or compromise while reducing mission risk from dependence on cyber resources.  

Recovery should therefore be configuration-specific and evidence-based. Restoring connectivity does not prove that credentials, software, data, and control paths are trustworthy again. Decisionmakers should know which configuration is returning, what evidence supports renewed trust, what limitations remain, and who accepts the residual risk.  

Measure the Denial Capability  

Three measures can turn these concepts into a deterrence capability that leaders can evaluate. The first is containment time: how quickly can the defender restrict the suspected dependency? The second is mission effect preserved during isolation: what essential service remains available? The third is trusted recovery time: how long before access and authority can be restored on a defensible evidentiary basis?  

These measures matter across alliances as well as inside a single operator. President Trump’s June national cybersecurity memorandum calls for key military and intelligence missions to remain viable in contested cyber environments and for greater coordination across agencies, industry, and international partners. Shared cloud services, communications, logistics systems, and identity providers make one partner’s isolation decision capable of affecting another partner’s readiness.  

No serious cyber strategy can promise that capable adversaries will never penetrate important systems. The stronger deterrence signal is that penetration does not guarantee strategic leverage. A defender that can detect a foothold, sever the dangerous dependency without surrendering its essential mission, and restore trust for evidence makes access less valuable in the first place.  

A decoy can reveal that an adversary is inside. Deterrence begins when the defender can take that access away without taking itself offline. 

Burak Oktenli is an independent researcher based in Washington, DC, conducting research on deterrence, cyber resilience, assurance, and machine authority in autonomous systems. His analysis has appeared in Global Security Review, RealClearDefense, the Modern War Institute at West Point, RUSI, and Articles of War. Views expressed in this article are the author’s own.

Cyber Deterrence Begins with the Power to Disconnect was originally published on Global Security Review.

Artificial intelligence is quietly redrawing the front lines of national security, and the United States’ Defense Industrial Base (DIB) now sits in the crosshairs of a new style of conflict. What once looked like routine cyber incidents now appears as a persistent campaign that fuses digital intrusion, supply chain manipulation, and targeted information operations into a single pressure system aimed at the factories, laboratories, and logistics networks that keep American forces armed and ready.

When most Americans imagine national defense, they focus on the uniformed force and the platforms that dominate headlines. Behind that visible edge of power stands a sprawling, largely private ecosystem of designers, manufacturers, software firms, logistics providers, depots, and research institutions that convert national intent into fielded combat capability. Federal regulation describes this DIB as the combined government and private sector industrial complex that can research, develop, produce, and sustain weapon systems and their components at the scale required for military operations.

That chain is far more diverse than the marquee names on major defense contracts. It includes niche machine shops, regional logistics firms, university laboratories, software providers, and government facilities that share technical data and workflow information through a constant flow of digital exchanges. Design files, bills of material, maintenance records, and operational telemetry form the nervous system of this industrial organism. When the system is healthy, the nation can surge production and sustain readiness under stress. When it is strained, deterrence becomes much harder to demonstrate in a crisis.

This industrial nervous system is absorbing AI at high speed, often with limited guardrails. Across the DIB, machine learning supports predictive maintenance, defect detection, demand forecasting, scheduling, and supply chain visibility. Corporate leadership often has only a surface-level understanding of how thinking machines can erode corporate identity, trust, and reputation quietly, yet catastrophically.

For the strategist, this adaptation expands the terrain adversaries can probe. Industry surveys in the World Economic Forum’s Global Cybersecurity Outlook 2026 report show that executives see AI as the primary driver of change in cybersecurity and regard AI-related vulnerabilities as the fastest-growing category of cyber risk. Organizations recognize that their attack surface is expanding as AI tools move into sensitive workflows, yet governance practices and security controls often lag deployment.

Hybrid campaigns against the DIB rarely announce themselves through spectacular outages. Security officials describe a patient pattern that begins with access and aims for endurance rather than immediate disruption. Once inside a network or supplier ecosystem, adversary teams map production rhythms, supplier dependencies, identity systems, and critical decision points where even modest delays can ripple across schedules and inventories. AI has become a force multiplier

for this reconnaissance, mining engineering files, and communications at machine speed and surfacing patterns that would take human analysts far longer to identify.

The contest is unfolding below the threshold of open conflict. Strategic competitors apply pressure through cyber intrusion, intellectual property theft, data manipulation, counterfeit insertion, and narrative operations designed to erode confidence in the U.S. capacity to arm itself and its allies. Hybrid pressure often manifests as subtle slowdown, rising cost, and creeping doubt rather than overt sabotage.

The stakes for deterrence are significant. Modern deterrence depends not only on visible platforms and war plans but also on an adversary’s belief that the United States can sustain forces and generate combat power at speed. If competitors conclude that AI-enabled systems in the DIB are brittle or easily manipulated, the credibility of deterrence weakens. The 2026 National Defense Strategy emphasizes strengthening the industrial base as a pillar of national power, tying cyber resilience directly to the ability to deter and wage war.

Pentagon planners increasingly describe cyber activity against the DIB as a persistent campaign environment. Adversaries can pursue national security effects through intrusions into private networks and business processes that support critical programs.

The technical attack surface is broad because the DIB operates as a tightly coupled system. Engineering teams collaborate across companies, moving design data through shared tools and platforms. Components flow through multiple tiers of suppliers before the final assembly. A single compromised vendor can offer adversaries access to multiple programs. Hybrid campaigns exploit these seams.

The effect of these operations extends beyond physical outcomes. Manipulated or stolen data can become material for influence narratives that frame U.S. production as unreliable or compromised. Production delays caused by cyber incidents can be amplified to sow doubt among allies about American capacity.

Adversary teams also benefit from repetition. They have treated the DIB as a learning environment, refining tradecraft through continuous operations. Hybrid coercion in this context is deliberate and calibrated, remaining below the threshold that would trigger a national response while still generating uncertainty and doubt.

Emerging evidence suggests that agentic AI systems are beginning to automate portions of this intrusion lifecycle. Automated chains reduce the cost of probing many suppliers and narrow the window for defenders to respond. For smaller firms with limited security resources, the pressure can quickly become unmanageable.

Security experts argue that defending against such campaigns will require disciplined governance and architecture. Pre-deployment AI security assessment and continuous validation must become standard practice, covering model selection, data provenance, access control, logging, and periodic reevaluation.

Architectural choices will also play a critical role. Zero trust principles and micro segmentation can reduce the impact of intrusions by limiting movement across networks. For DIB operators, this means treating segmentation and strong identity controls as the default architecture.

The industrial context around these cyber efforts is already volatile. The aerospace and defense sector is managing supply chain disruptions and resource constraints while trying to scale production. This environment amplifies the effectiveness of cyber sabotage and influence operations because attacks land in systems that are already stressed.

The strategic logic behind adversary focus on the DIB is clear. By degrading confidence and throughput, competitors can undercut deterrence without open conflict. When intrusions are contained and production continues under pressure, the calculus of hybrid aggression shifts.

Hybrid warfare against the DIB aims to bend trust in the systems that generate combat power. If those trust systems are distorted through advanced technologies and coordinated information operations, the result is slower decisions, higher costs, delayed output, and a quieter form of coercion.

Mr. Greg Sharpe is a Fellow and the director of Communications and Marketing for the National Institute for Deterrence Studies and the Managing Design Editor for the Global Security Review. Greg has over 35 years of military, federal civilian, and defense contractor experience in the fields of database development, digital marketing & analytics, technology use case exploration and assessment, and as a USAF doctrine outreach and engagement analyst. The views of the author are his own.

Quiet Warfare: Bending Data and Perceptions in the Defense Industrial Base was originally published on Global Security Review.