Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Enterprise Cyber-Physical Damage and Insurance Coverage Coordination

Abstract. Enterprise cyber-physical incidents can couple a malicious digital intervention with damaged machinery, industrial shutdown, lost output, environmental consequences, bodily injury, and data impairment. Yet insurance programs are organized by insuring agreements, insured assets, causes of loss, financial interests, and contract-defined periods. A single engineering sequence can therefore generate multiple claims under property, cyber, business interruption, machinery, general liability, and specialty policies. This monograph examines that coordination problem using industrial-control guidance, documented incidents, insurance-market endorsements, insurer-supervisory standards, judicial decisions, and actuarial literature. It distinguishes established tangible damage from operational disruption and digital impairment; reconstructs the chain from software compromise to physical effect; analyzes coverage grants, exclusions, write-backs, retentions, waiting periods, sublimits, and contract priority; and evaluates international supervisory context. A reproducible synthetic loss scenario demonstrates how different contractual conditions alter the recoverable amount. The original contribution is an evidence-based analytical synthesis of technical causation, loss classification, and contractual response. Neither a newly authored theoretical model nor a universally applicable insurance-coverage conclusion is claimed.

Keywords: Cyber-physical damage; operational technology; industrial control systems; cyber insurance; property insurance; physical loss; business interruption; cyber exclusions; silent cyber; claims coordination; liability; insurance contract interpretation; loss aggregation; insurance supervision; insurance accumulation.

Introduction

A compromised controller may create unexpected pressure, heat, speed, chemical composition, or movement. A digital intrusion may alternatively cause a plant to shut down safely, interrupt a utility, corrupt files, or prevent operators from observing a process. The engineering consequences are different, and the associated insurance rights may be different again. Enterprise cyber-physical loss presents a problem at the intersection of industrial engineering, forensic causation, property valuation, business income, liability, policy interpretation, and underwriting accumulation.

This research proceeds from three connected questions. First, what credible evidence links malicious digital conduct to a change in the physical operation of equipment? Second, which economically distinct consequences arise from that sequence, and how can they be measured without duplicate counting? Third, how do the actual insurance contracts allocate each consequence through their insuring agreements, exclusions, endorsements, deductibles, time restrictions, and limits? A technically convincing description of a cyberattack cannot establish an insurer's liability on its own. A policy description cannot determine what physically occurred. An announced claim and a final indemnity payment are different financial observations.

The investigation encompasses manufacturing facilities, electricity and water operators, industrial supply chains, controlled buildings, logistics infrastructure, and enterprises dependent on shared operational platforms. It is international in scientific scope, but legal conclusions are traced to specific instruments and jurisdictions. The examples are disclosed as hypothetical; no actual enterprise's confidential program is represented. Sources include the final NIST OT security guidance, publicly accessible incident investigations, original insurance-supervisory documents, published court opinions, and peer-reviewed actuarial and engineering research.

Research method and evidentiary hierarchy

Primary technical evidence includes known controller functions, network and system records, process-historian data, operating logs, forensic images, hardware inspections, and safety-system behavior. The published case studies considered here summarize such evidence to differing degrees; they do not replace access to the plant's own complete engineering record. Primary contractual evidence is the issued policy with declarations, endorsements, definitions, exclusions, relevant schedules, and applicable law. A market model wording is evidence about available drafting practice, not proof that a company purchased it. Judicial findings are interpreted with respect to the particular contract, procedural posture, and facts. Regulatory guidance is cited for regulatory expectations within its jurisdiction, not as a universal private-law rule.

The monograph uses a source-to-claim audit: each linked document supports the sentence in which it is invoked; every linked external research document receives a bibliographic entry and a record in EVIDENCE.json. Peer-reviewed sources accessed through publisher metadata or an abstract are used only for claims supported at that level. The quantitative exercise contains transparent arithmetic rather than an empirical estimation. Historical and current statements are anchored to November 16, 2026. The NIST OT publications index identifies Revision 4 of SP 800-82 as an initial public draft issued in September 2026; Revision 3 remains the last final edition at the research cutoff.

The original synthesis treats the engineering event, the economic loss, and the insured claim as three observable structures requiring a documented mapping between them. This is an analytical distinction applied to established engineering, accounting, and insurance concepts. It is not introduced as a new doctrine or theoretical system.

Research Object, Definitions, and Historical Foundations

Industrial cyber-physical damage

Operational technology comprises programmable equipment and connected systems that monitor or directly influence physical environments and industrial processes. Programmable logic controllers, distributed control systems, supervisory control and data acquisition systems, protection relays, building automation, and safety instrumented systems are relevant examples. Stouffer et al.'s NIST SP 800-82 Rev. 3 emphasizes distinctive reliability, safety, and availability requirements. This engineering feature is crucial for insurance because an altered program can affect assets beyond the computer on which it executes.

For research classification, cyber-physical damage is tangible impairment, destruction, or harmful material alteration arising through cyber-origin manipulation of a physical system. The definition identifies an empirical class of loss; it does not displace any contract-specific legal meaning of "physical loss or damage." A contract might expressly include data corruption in its definition, or it might exclude certain electronic events despite associated physical loss. The analyst must preserve both the technical and contractual vocabulary.

A rigorous incident description distinguishes at least four consequences. Physical property can be damaged or destroyed. Operators may lose control or protection while the machinery remains intact. Production may be suspended as a safety precaution without demonstrated damage. Digital information or computer functionality can be impaired, and whether that impairment constitutes insured physical damage may depend on policy text and applicable law. MITRE's damage-to-property classification and separate productivity and revenue impact reflect the substantive distinction between equipment destruction and interrupted output. These are descriptions of attacker outcomes, not coverage grants.

Safety instrumented systems perform a specially important protective function by directing processes toward a safe state. The MITRE industrial-control technique catalog identifies loss of protection and loss of safety separately from property damage. A protective shutdown can demonstrate that a system responded to an abnormal condition; it does not automatically demonstrate that tangible plant assets were injured. Conversely, compromise of a safeguard may increase the probability of damage that has not yet occurred. An insurer and an engineer need records sufficient to distinguish a potential hazard from an observed physical outcome.

The historical emergence of coverage ambiguity

Industrial systems became more digitally integrated while traditional commercial insurance maintained distinct policies for buildings, machinery, stocks, business income, third-party injury, and financial losses. Some earlier contracts used broad all-risk language while scarcely referring to cyber causation. Such policies could create non-affirmative or "silent cyber" exposures, where cyber losses might fit a coverage grant without explicit pricing or exclusion. This concern extends across property and casualty lines; its importance rises when a single digital event affects several insured locations or multiple policy classes.

The Bank of England Prudential Regulation Authority's Cyber Insurance Underwriting Risk statement calls for active identification and management of non-affirmative exposure. EIOPA's supervisory statement on silent cyber addresses the same issue in the European supervisory setting. These documents demonstrate that coverage ambiguity is an issue of policy clarity, risk measurement, and potential accumulation. They do not themselves decide a particular claim.

The Lloyd's Market Association's 2019 Property and Marine Cyber Clauses bulletin reports publication of illustrative model endorsements and exclusions, including LMA5400 and LMA5401. The association expressly states that underwriters may change the model terms. Accordingly, references to an LMA clause number describe an available drafting family; a valid coverage analysis must obtain the precise wording and its insertion into the executed policy. There is no single universal "property cyber exclusion."

The enterprise economic loss record

The initial engineering incident frequently generates a vector of losses: physical repair and replacement, damaged inventory, business income, extra expense, information-system restoration, forensic investigation, interruption at dependent locations, contractual disputes, and claims for injury or environmental damage. These amounts are measured at different stages. Gross economic loss must be distinguished from covered loss before retention, from an insurer's case reserve, from an agreed settlement, and from final cash paid. Insurance limits are maxima subject to conditions rather than substitutes for these observations.

The enterprise should assign the same event identifier to every component while separately recording owner or legal entity, insured interest, location, chronology, cause, documents, valuation method, gross amount, claimed amount, potential coverage, and actual payment. Such separation prevents a lost-profit item from being counted both as a stand-alone economic loss and as an expense within property damage. It also enables later accounting for salvage, recovery from responsible suppliers, and subrogation. These are established loss-adjustment and accounting principles applied to an unusually intertwined event.

Cyber Control-System Causation and Incident Evidence

Reconstructing the cyber-to-physical pathway

An engineering investigation links the initial compromise to the alleged loss through intermediate states. Typical stages include entry through remote access or credentials, movement from enterprise IT to an OT network, use of engineering workstations or legitimate management tools, changed controller logic or commands, alteration of measured or controlled parameters, interaction with physical safeguards, actual physical outcomes, and restoration or emergency shutdown. The sequence is documented with synchronized clocks, logs, configuration snapshots, plant historian records, independent sensor measurements, maintenance history, and material inspection.

Causation requires the examination of alternatives. A mechanical failure may have preceded the digital alarm. An operator might have initiated a shutdown; a safety interlock might have prevented the feared event; or malicious commands might have been accepted but never executed. Source preservation is essential because rapid restoration can overwrite the evidence required to distinguish these scenarios. A forensic account should state what was observed, what was reconstructed from circumstantial data, and what remains uncertain.

Zhang et al. (2021), in their survey of industrial cyber-physical protection, review attack detection and secure estimation in systems involving control, communication, and computing. The technical literature clarifies the difference between detecting abnormal messages and establishing a process consequence. For the insurer, this distinction affects both event classification and the evidence supporting a repair, interruption, or liability payment. It also prevents an analyst from treating proof of attacker capability as proof that the capability caused physical damage.

Ukraine electricity incidents: interrupted physical service

The December 2015 coordinated cyber intrusion into Ukrainian electricity distribution systems produced power interruptions through unauthorized access to operational control functionality. Lee, Assante, and Conway's analysis of the Ukrainian power grid describes compromise, manipulation of operational technology, and restoration difficulties. This case establishes loss of service and production-relevant operational harm. It does not support an automatic inference that every affected customer experienced direct physical injury to insured property.

The subsequent E-ISAC/SANS Modular ICS Malware study distinguishes the 2015 incidents from malware associated with the 2016 attack and industrial communication protocols. For insurance purposes, a substation control event, an interruption of power supply, a damaged transformer, and a customer's lost profit must be examined as separate facts. A policy might grant utility-service interruption, dependent-property coverage, or specific non-damage interruption, but eligibility cannot be inferred from an outage headline.

TRITON and safety systems

Mandiant's first-hand December 2017 TRITON incident report describes malicious code designed to interact with Triconex safety instrumented system controllers. The researchers assessed with moderate confidence that the intruder was developing capability for physical damage; the observed result included an inadvertent operational shutdown. The investigators' language is important: capability for damage, actual shutdown, and actual severe physical destruction are distinct findings. The public case does not establish that the anticipated physical destruction happened.

The follow-up TRITON and TriStation analysis explains why industrial-protocol and device-specific knowledge matters. A potentially hazardous attack may produce a loss consisting principally of lost output, response expense, and safety-system verification. A later occurrence that causes ruptured equipment or contamination would produce additional property and liability questions. Insurance analysis should follow the observed outcome and the issued terms, not the severity of the threat actor's apparent intention.

The German steelworks report

The German industrial-security literature discusses a 2014 steelworks incident in which a malicious intrusion reportedly proceeded from the business network into production control and was followed by a failed controlled shutdown and substantial physical damage. The GRS report Industrial Control-System Threat Situation summarizes the account while acknowledging uncertainties about the responsible actor and exact malware. It supports an account of reported material damage and a plausible IT-to-OT causal sequence. The public report does not disclose the full engineering adjustment file, precise final loss valuation, issued insurance program, or settlement record.

The difference among these cases is itself a scientific result. The category "industrial cyber incident" encompasses realized property destruction, loss of control, protective shutdown, and service interruption. Aggregating them as identical physical damage observations would distort both the numerator and the denominator of any insurance severity study. The evidence also suffers selection effects: public cases are not a representative statistical sample of industrial policyholders.

Current exposure and operational evidence

CISA's September 2024 warning on exposed OT devices describes continued exploitation of internet-facing operational systems, including water-sector assets, through commonplace access weaknesses. The organizational NIST Cybersecurity Framework 2.0 provides a governance taxonomy that can align cybersecurity and enterprise risk functions. Engineering precautions include segmented networks, managed remote access, secure identity controls, asset inventories, independent safety layers, resilient backups, and rehearsed recovery procedures. These influence exposure and forensic quality. They become contractual insurance requirements only when the policy or proposal terms give them that status.

The insurance implication arises before a loss. Asset inventories and process-dependency maps establish which cyber events could generate damage to tangible property and which would instead interrupt output. A well-documented safety architecture also makes post-incident root-cause reconstruction more reliable. A cyber maturity score alone does not establish the financial severity of a failed industrial process.

Physical Damage, Contract Triggers, and Exclusions

Property damage and the wording of the insuring clause

A first-party property claim begins with the grant of coverage: covered property, insured interest, insured location, cause of loss, limits, and the operative definition of physical loss or damage. A malicious control command can lead to a seized motor, overheated furnace, ruptured tank, destroyed stock, or contaminated material. Such effects can be documented through metallurgical examination, engineering inspection, process logs, and repair or replacement estimates. Whether the damage is insured depends on the contract. Physical evidence is essential to prove physical alteration, yet the engineering fact of damage does not override a policy exclusion for certain digitally initiated causes.

The exact text of cyber exclusions and ensuing-loss or write-back provisions must be reconstructed. Some endorsements distinguish the compromise of an electronic system from ensuing fire or explosion; others address electronic data, malicious acts, malfunction, or subsequent tangible damage differently. A loss may also involve concurrent mechanical faults or human intervention. Contractual causation, anti-concurrent-causation provisions, efficient-proximate-cause doctrines where applicable, and the wording of a specific ensuing-loss exception can therefore matter. The mere existence of a cyberattack does not resolve which exclusions apply.

The Lloyd's Market Association's property and marine clause package offers documented examples of explicit cyber wording, including the LMA5400 and LMA5401 forms. The LMA describes these as illustrative model clauses, which participating underwriters may amend. For a particular enterprise, the relevant evidence is the exact executed endorsement and the policy provisions it modifies. Market shorthand such as "affirmative cyber" or "cyber-excluded property" cannot replace that reading.

Property valuation generates an additional layer. The damaged asset may be insured at replacement cost, actual cash value, an agreed value, or another contractual measure. The insurer may dispute whether replacement is technically necessary, whether a repair entails betterment, or whether the asset's ordinary deterioration contributed to the loss. Restoration of industrial equipment can require engineering certification and testing even when replacement parts are readily available. The economic effect of lost production during testing is considered separately under time-element wording.

Business interruption and the requirement of physical loss

A traditional damage-dependent business-income grant may require covered physical loss or damage at insured premises and interruption caused by that condition, subject to a defined restoration period. A plant whose controllers are compromised but whose equipment remains physically functional may face real lost income while failing a particular property-policy condition. The analysis must distinguish the economic reality of disruption from the policy's specified trigger. The start and end of the period of restoration, waiting period, saved variable costs, production backlog, and adjusted revenue forecasts also affect indemnity.

Standalone cyber insurance can offer network business interruption associated with system failure or security failure, potentially without tangible damage. Its own definitions may distinguish security failure, accidental system failure, operational technology, voluntary shutdown, and third-party service-provider outage. Some forms exclude bodily injury and damage to tangible property, and some specify conditions for interruption of industrial systems. Cyber insurance can therefore respond to a network-only loss that does not meet property physical-damage requirements, while property coverage may respond to material repair and damage-dependent profit loss. The actual combination can create overlaps or gaps.

A safety shutdown illustrates the distinction. The operator of a hazardous plant may suspend production to avoid an explosion after evidence of compromised protective logic. This may be a justified safety decision and a significant business expense. For insurance, the relevant questions include whether a qualifying insured security event occurred, whether shutdown expenses fall within a named grant, whether a non-damage interruption extension applies, and whether any waiting period or prevention-of-access clause controls. Claim evidence should identify who ordered shutdown, on what engineering grounds, and when safe restart became possible.

Computer impairment and decisions of courts

Courts have considered the boundary between digital impairment and direct physical damage in particular insurance wording. In National Ink and Stitch v. State Auto, the United States District Court for the District of Maryland held in January 2020 that ransomware-related impairment of computer functionality qualified as physical loss or damage for the contract and factual record at issue. The opinion supports that specific result. It does not establish that every digital attack must satisfy every physical-damage clause.

The New Jersey Appellate Division's 2023 judgment in Merck & Co. v. ACE American concerned the NotPetya attack and property policies whose definitions expressly addressed computer data, coding, program, and software destruction or corruption. The court held that the hostile-or-warlike-action exclusion, as historically drafted, did not bar the claimed loss on the record before it. The opinion distinguishes a specific contract's treatment of digital impairment and nonphysical computer-system interruption. It neither declares data corruption to be material destruction in all contracts nor forecloses differently worded modern war or state-backed-cyber exclusions.

The methodological consequence is that a legal comparison must identify the actual edition of the insuring clause, exclusion, endorsements, governing jurisdiction, court level, and procedural posture. Legal outcomes should be expressed as decisions on particular instruments rather than transformed into general scientific claims that "cyber damage is covered" or "cyber damage is excluded." In cross-border programs, a single attack can produce claims subject to different domestic policy terms and legal systems.

Machinery breakdown and specialist expenses

Equipment-breakdown and machinery policies can have different triggers from general property insurance. A digitally modified control instruction might contribute to sudden electrical breakdown, abnormal rotational stress, or destructive pressure. Claims analysis must establish the physical breakdown mechanism and assess whether wear, gradual deterioration, electrical arcing, design defect, or digital causation is included or excluded by the actual form. A controller's failure and damage to the controlled machine are separately recorded; their financial valuation and retention may differ.

Cyber response grants may address digital forensics, system rebuild, data restoration, malware removal, breach-response specialists, and incident consultants. Physical-property claims may address inspection, repair, debris removal, and testing. Extra expense and mitigation extensions can cover emergency rental equipment, expedited transport, temporary production, or overtime when the specified contractual tests are met. Expenses incurred to modernize an aging plant or improve cybersecurity during restoration may need separation from reinstatement of the prior function. A total restoration invoice should be allocated by underlying service and purpose rather than submitted identically to every insurer.

Third-party injury, pollution, and financial responsibility

An industrial cyber-physical event may cause injury to visitors, customers, neighboring facilities, or the environment, as well as loss to the enterprise itself. General liability, product liability, environmental liability, employers liability, professional liability, and contractual indemnity arrangements can each be relevant. These are different legal and insuring interests: repairing the insured's own machine is not the same as defending a claim alleging damage to a neighbor's premises. Occurrence, claims-made, and claims-made-and-reported wordings may respond differently to the same factual sequence, particularly where notice and reporting cross policy periods.

Environmental cleanup is especially contract-dependent. A cyber-triggered industrial discharge may generate first-party remediation costs, third-party property damage, statutory obligations, and potential fines. Pollution exclusions, sudden-and-accidental provisions, specialist environmental grants, and local laws of insurability may produce different outcomes. It is incorrect to equate every remediation charge with insured indemnity or to treat a fine and civil compensatory damage as the same insured item. Each component needs a distinct contractual and legal finding.

Hostile operations, war exclusions, and attribution

The category of state-backed or hostile cyber activity introduces a combination of attribution uncertainty and contract drafting. The Lloyd's war and NCBR underwriting requirements document market expectations and references to state-backed cyber exclusions. Their regulatory-market role must be distinguished from an insured's actual wordings. A hostile or war-related exclusion may turn on definitions, exclusions from exclusions, territorial consequences, attribution standards, and the treatment of civilian infrastructure.

Merck demonstrates that a court may interpret an older hostile-action clause narrowly in its contractual setting. More recent specialist cyber exclusions have distinct drafting. Proof that a security analyst suspects state sponsorship does not establish an insurance contract's definition of state-backed activity, and public attribution may differ from the contract's decision mechanism. A careful loss file records the precise exclusion, the supporting evidence, the reason for reserving or denying rights, and the date of any final determination.

Multi-Policy Claims Coordination and Financial Allocation

The relationship among the event, the loss, and the contract

Insurance coordination begins with one stable technical incident record and a detailed ledger of economically distinct consequences. A single incident can create damage to property, operational downtime, response costs, third-party injury, and contingent interruption at locations outside the affected plant. Each consequence is then evaluated separately under the relevant policy. A contractual occurrence definition can aggregate several technical effects; separate contracts can define different occurrences. An enterprise-level analytical event identifier therefore assists traceability but does not impose one common legal trigger.

At the level of an individual claim, the record should contain the event identifier; asset and legal-entity identity; insured interest and geographical location; time and cause of loss; technical supporting documents; gross economic value; relevant insuring agreement; exclusions and conditions; eligible amount before retention; attachment point, deductible, waiting period, sublimit, and aggregate; claimed and reserved amounts; insurer communications; payments, salvage, and recoveries; and outstanding disputes. This permits the full accounting chain to be reconstructed. Unrelated expenses should not be grouped only because they appear on the same invoice.

The PRA's non-affirmative cyber supervisory framework and the IAIS's underwriting challenges report identify the risk of unrecognized cyber exposure in multiple insurance classes. For a corporate insured, an inventory of property and cyber coverage serves both loss management and the ex ante measurement of residual risk. For insurers and reinsurers, it can reveal concentrations obscured by separate line-of-business reporting.

Executed policies and legal-entity boundaries

The program may contain a global property master policy, local admitted policies, a standalone cyber program, equipment breakdown, general liability, environmental liability, captive retentions, and excess layers. A diagram of nominal limits is insufficient to determine which policy responds. Analysts require the issued policy and endorsements, declarations, covered locations, applicable law, named insureds, currency, local admitted requirements, applicable definitions, retroactive dates, underlying exhaustion requirements, and other-insurance clauses. Contract hierarchy can depend on the particular language and governing law.

A multinational parent may own production equipment at one site, lease equipment at another, and purchase utilities from independent operators. Contractual indemnities from vendors, suppliers, and service providers can affect ultimate economic burden without expanding insurance coverage. The local insured's recoverable interest may differ from the group's consolidated financial loss. The policy period may also matter: the intrusion, subsequent damage, discovery, third-party complaint, and claim reporting can occur on distinct dates. Occurrence-based contracts, claims-made provisions, related-act clauses, and extended reporting periods must be analyzed as actually written.

The cross-border implication is a demand for precision in coverage records. There is no generally applicable rule that one global master policy automatically pays all losses suffered by every subsidiary wherever located. A company must identify the contracting insurer, insured legal entity, loss location, and national regulatory conditions before presenting a consolidated group recovery.

A transparent synthetic claims exercise

Consider a synthetic manufacturer subjected to an unauthorized control command. The resulting disturbance damages machines and inventory, stops one production line, and triggers separate security-response expenses and network interruption. A third party alleges injury or property damage. The components below are nonoverlapping enterprise economic loss categories, valued in USD millions. They are assumptions for illustrating contractual mechanics, not observations drawn from a company or market dataset.

Loss component Gross economic loss (USD millions) Hypothetical coverage pathway
Damaged machinery and inventory 60 Property grant, subject to covered cause and exclusions
Damage-dependent loss of production gross profit 20 Property business income, if physical-damage trigger is satisfied
Cyber investigation and digital restoration 8 Cyber incident response and restoration grant
Separate network-only interruption 12 Cyber non-damage business interruption, subject to waiting period
Third-party property damage or injury liability 15 Liability terms, exclusions, defenses, and limits
Total gross economic loss 115 Enterprise loss before deductibles, limitations, and recovery

Assume that the physical property and damage-dependent income components jointly qualify for the property policy, yielding USD 80 million before a single USD 5 million deductible; and that the policy has one USD 50 million combined limit. The hypothetical property payment equals min(50, 80 − 5) = USD 50 million. This assumes one shared deductible, no smaller sublimit, and no other restricting condition. Real policies can apply different arrangements.

Assume independently that USD 8 million of cyber investigation/restoration expense is within the cyber grant, with a USD 1 million monetary retention, and that USD 3 million of the USD 12 million network-only interruption falls outside an insured waiting-period adjustment. The hypothetical eligible total is (8 − 1) + (12 − 3) = USD 16 million. If one shared cyber limit of USD 15 million applies, the payment is USD 15 million. There is no overlapping income item in this original table.

Assume the USD 15 million third-party liability component is payable to the extent of USD 10 million in one contractual scenario, after all applicable terms and limits. In a second scenario, a valid exclusion bars that liability component, reducing the payment to zero. The first scenario produces total payments of USD 50 million + USD 15 million + USD 10 million = USD 75 million and residual economic loss of USD 40 million. The second produces USD 65 million paid and USD 50 million residual. No recovery from suppliers, salvage, or other sources is assumed. The arithmetic identity is:

Residual loss=Gross economic loss−Insurer indemnity−Other realized recoveries.\mathrm{Residual\ loss} = \mathrm{Gross\ economic\ loss} - \mathrm{Insurer\ indemnity} - \mathrm{Other\ realized\ recoveries}.

The example demonstrates the sensitivity of recovery to grants, terms, and exclusions. It is not a claim about the average extent of industrial cyber coverage, a recommendation to buy any specific amount of insurance, or an actual claims adjustment. The loss table is deliberately constructed to avoid duplicate counting, allowing the contractual assumptions to be examined independently.

Overlapping grants and double counting

Now suppose one USD 6 million interruption component satisfies the factual description of both damage-dependent property business income and cyber network interruption. The enterprise has one USD 6 million economic loss even if both grants are potentially implicated. The proper accounting records the amount once and then evaluates coordination among contracts. "Other insurance," contribution, excess, priority, noncumulation, and related-event provisions can determine how more than one potentially responsive insurer shares the burden. The legal outcome depends on operative policy language and applicable law, including judicial treatment of conflicting clauses.

The audit requirement is to prevent duplicate collection of indemnity for the same actual economic loss while retaining separate contractual payments for distinguishable services or exposures. A defense-expense benefit, a specific incident-response service, and the actual lost gross profit are not necessarily duplicate loss categories. Equally, a reimbursement labeled "extra expense" should not be paid again as the same cost under another invoice category without a valid independent basis. Subrogation and third-party recoveries can alter the final financial distribution after initial insurance payments.

Deductibles, waiting periods, sublimits, and exhaustion

Deductibles, self-insured retentions, and waiting periods need separate treatment because each can apply to a different measure. A monetary deductible may reduce the eligible property amount; a retained layer may govern the insurer's attachment; and a time-based waiting period can leave the first hours of interruption uninsured. An event sublimit may govern several coverage extensions collectively, while an annual aggregate may affect later events. At a corporate group level, a single common cyber occurrence could impair multiple sites, each subject to separate property deductibles and location schedules.

Excess coverage does not become payable merely because gross group loss exceeds the nominal primary limit. The insured must establish the attachment, eligible underlying loss, exhaustion requirement, applicable sublimits, and provisions concerning payment by the underlying insurer. Policy towers can treat a digital event differently at successive layers. A useful pre-loss exercise tests identical, explicit incident narratives against every underlying and excess form rather than assuming uniformity.

Claims governance and reservation of rights

An enterprise's claims team needs to notify potentially responsive insurers as required by their respective policies while avoiding inconsistent descriptions of the same physical event. A prompt notification should identify known facts and express uncertain points as uncertain. Later expert findings can alter the loss allocation. A reservation-of-rights letter records an insurer's stated position at a particular stage; it is not equivalent to a final judgment or payment. The claim register should preserve revised assessments and the provenance of each change.

Effective governance joins OT engineering, finance, legal, risk management, brokers, loss adjusters, and insurers through controlled records. An engineering report should be intelligible to contract analysts, and a coverage opinion should identify the technical facts on which it depends. This linkage is the principal operational advantage of a coordinated claims process: it makes unresolved questions explicit and prevents loss components from disappearing between organizational teams.

International Comparison, Measurement, and Uncertainty

The United Kingdom and European Union supervisory contexts

A central development in non-affirmative cyber regulation has been the explicit recognition that cyber-origin losses can accumulate inside classes of insurance originally underwritten as property or liability. The Prudential Regulation Authority's SS4/17 cyber underwriting statement addresses UK insurers' identification, quantification, management, and monitoring of cyber underwriting exposure. The version displayed by the Bank of England was published November 15, 2024, with an effective date of December 31, 2024. It provides supervisory expectations, not contract language automatically incorporated into every commercial policy.

EIOPA's 2022 statement on non-affirmative cyber exposure similarly focuses on the need for national competent authorities to address silent cyber risk across traditional insurance lines. The statement belongs to the European supervisory framework and cannot be read as a binding interpretation of an individual policy exclusion. It supports a scientific conclusion that property and cyber portfolios cannot be analyzed in isolation when a common originating event could produce claims across both.

Legal and operational obligations also coexist. The European Union's NIS2 Directive establishes cybersecurity risk-management and incident-reporting duties for entities within its scope, subject to national transposition and the directive's provisions. The duties may influence enterprise security architecture, reporting, and governance. They are analytically distinct from the contractual triggers for physical damage and business interruption. Compliance with regulatory cyber standards does not, by itself, establish insurance recovery, and a policyholder's contractual obligations must be identified in the issued form.

The IAIS's 2025 Global Insurance Market Report discusses cyber-risk evolution, underwriting exposure, and challenges of insurability and clarity. The IAIS and Financial Stability Institute's June 2026 cyber insurance market note likewise points to continued limitations in data, pricing, coverage, and accumulation. Both are institutional analyses, not direct estimates of the proportion of industrial control incidents that caused insured material destruction.

Supervisory objectives and contractual outcomes

Regulators seek adequate insurer governance, reasonable understanding of exposures, and resilient underwriting. Underwriters may seek to clarify exclusions and affirmatively price selected cover. Corporate policyholders seek reliable indemnity after specified losses. These related objectives do not yield identical outcomes. A model exclusion can reduce an insurer's unintended accumulation while shifting retained cyber-physical exposure to the corporate balance sheet. A broad affirmative extension can enhance coverage predictability while raising premium, aggregation, or capacity concerns.

The Geneva Association's Cyber Risk Accumulation report examines accumulation and insurability, including cyber losses that affect physical property or bodily injury. It highlights why scenario-based understanding matters beyond privacy breaches or isolated ransomware events. The joint treatment of property, cyber, contingent, and liability lines is therefore a substantive underwriting need. The report cannot establish the response of any unknown enterprise contract, because the response is conditional on terms and local law.

Lloyd's market requirements concerning war-related wording and the LMA's illustrative clauses form another regulatory-market layer. Their function is to promote explicit treatment of particular risks and more consistent contracting. A market-wide initiative is evidence of institutional recognition of coverage ambiguity. Whether it succeeds in removing uncertainty from an individual claim is an empirical and legal question that can only be tested against actual forms, coverage decisions, and later dispute outcomes.

Insurance data and the problem of denominators

The National Association of Insurance Commissioners' 2026 U.S. cyber insurance market report covers reporting year 2025 and reports approximately USD 9.46 billion in direct written cyber premium, including eligible alien surplus lines; domestic direct written premium was approximately USD 7.46 billion. The report identifies 56,323 cyber claims and roughly 5.84 million domestic policies in force. Those figures describe the reporting population captured in the regulatory market data. They do not determine how many attacks physically damaged industrial machinery.

Four denominators should be differentiated before empirical inference. Insured cyber claims are not equivalent to all reported cybersecurity incidents. Reported incidents are not equivalent to all industrial control breaches. Industrial control breaches are not equivalent to realized damage to tangible property. And damage to tangible property is not equivalent to a claim paid under a property or cyber contract. The NAIC's earlier cyber insurance market report describes the use of statutory and supplement data, demonstrating why premium and policy-count definitions matter. Comparisons must preserve reporting year, coverage class, source population, and coding conventions.

A meaningful claims severity analysis would classify observations by industrial sector, revenue and asset value, type of OT compromise, tangible outcome, restoration time, wordings and sublimits, and settlement outcome. Public reports rarely disclose all these variables. Average insured cyber claim severity should therefore not be interpreted as the expected insured physical destruction at a power plant or chemical facility without a justified, observed mapping between populations.

What the actuarial literature does—and does not—establish

Eling and Schnell's cyber-risk literature review identifies challenges concerning data, interdependency, and limits of risk modeling. Their synthesis supports caution in parameterizing low-frequency severe loss from incomplete observations. Eling, Jung, and Shim's quantile-regression analysis examines heterogeneity in observed cyber-risk outcomes, highlighting that explanatory relations need not have the same strength across a loss distribution. These findings are relevant to the design of a broader cyber-risk study, but their populations cannot be assumed to represent industrial cyber-physical damage claims.

Awiszus and colleagues' cyber insurance modeling survey examines quantitative representation and pricing, including interactions among insured risks. Established actuarial and network approaches can help identify correlations associated with shared software, cloud platforms, remote management, and service providers. Yet the severity of material damage is additionally shaped by process variables: temperature, pressure, safeguard configuration, maintenance condition, timing, and available operator intervention. Two plants compromised by the same software exploit can have dramatically different physical outcomes.

An illustrative contrast makes the dependency issue explicit. Suppose ten plants each face a hypothetical one-percent chance of one specified incident in a defined period. Under a simplifying assumption of full independence, the probability that all ten experience that incident is 0.0110=10−200.01^{10}=10^{-20}. If the plants instead depend on one shared service with a hypothetical one-percent probability of a failure that simultaneously affects every plant, the probability of simultaneous exposure becomes 0.01 under that differently specified scenario. The probabilities are invented solely to explain the effect of dependence; they are not observed industrial attack rates. The assumptions are intentionally incompatible and must not be confused with a calibrated forecast.

Propagation differs from physical severity

Shared dependence can produce correlated cyber access without identical physical damage. Compromise of an engineering service may propagate to one hundred controllers, while only a subset operates at conditions vulnerable to destructive commands. A safe shutdown in one plant and a rupture in another can arise from the same digital entry mechanism because local physical conditions and safeguards differ. Conversely, an outage of a common electricity supply may cause business interruption at many sites without physical damage at each. For coverage aggregation, one must model common-cause exposure, plant-specific outcome, and policy-specific trigger separately.

An insurance stress test should thus enumerate a plausible digital initiating event, the connections allowing it to spread, physical operations at the affected facilities, potential safety interventions, and the contract response to every resulting loss category. The test should show loss before insurance, potentially covered loss under each scenario, and the expected or possible contribution of retentions and limitations. Where the evidence cannot support a probability, the result should remain a stress scenario instead of a purportedly measured frequency distribution.

Comparative legal evidence and its limits

The judgments analyzed in this dossier concern United States policies. National Ink is a federal trial-court decision applying the law and wording before it; Merck is a New Jersey appellate decision addressing an identifiable program and exclusion. Together, they show that language governing computer impairment and hostile or warlike action can materially affect a coverage result. They do not supply an internationally uniform definition of physical damage or a universal rule on state-backed cyber events.

A valid cross-jurisdictional comparison would require the complete executed policy, governing law, translated or original local statutory texts, procedural status, later case treatment, and a verified factual record of the industrial event. Different legal systems may also regulate admitted coverage, insurable interest, sanctions, penalties, and insurance claims. An opinion about one jurisdiction should be stated as such, and conflicting or untested clauses should remain identified as matters of uncertainty. International generalization without these qualifications would exceed the evidence.

Source and measurement limitations

The strongest limitations are structural. Public technical incident reports commonly omit sensitive operational records, precise fault traces, and full material damage valuations. Corporate insurance wordings and settlements are often confidential. Observational cyber datasets can be weighted toward ransomware, data breach, and general network interruption. Peer-reviewed cyber-risk studies may therefore lack both the event classification and insurance-payment components required for causal claims about industrial physical damage.

The German steelworks evidence illustrates reported physical injury without publicly documented complete claims recovery. TRITON illustrates an extreme hazard and reported shutdown without proof of the hypothesized catastrophic destruction. Ukraine power disruptions illustrate operational harm without a general inference of insured damage at each affected facility. This is a coherent evidentiary result rather than an inconsistency: each event supports conclusions of a different kind and strength. The research consequently declines to estimate a global industrial cyber-physical insured-loss rate or the probability of coverage denial.

Reproducible enterprise case research would preserve a process timeline, system configurations, independent physical inspection, gross losses by component, actual executed coverage, event-specific exclusions, notices and decisions, and a final reconciliation of insurer payments and other recoveries. Each variable should carry a clear unit, source, timestamp, verification status, and access limitation. Without such records, apparent differences across cases may reflect reporting and contract selection rather than actual changes in engineering severity.

Integrated Findings and Research Agenda

First finding: event unity does not imply coverage unity

The technical root cause may be one malicious control action, while the resulting economic and insured consequences are many. The loss should retain a stable causal history even when insurance contracts aggregate or separate events differently. Forensic evidence establishes the operation of the control system; financial documentation establishes components of economic loss; contract evidence determines the coverage and retention applicable to each component. The three assessments are connected by traceable facts but remain distinct in their methods and results.

This finding explains why a dangerous cyber-physical event may generate substantial interruption expense with little established material damage, and why a less visible exploit can cause expensive equipment destruction. Classification by the label "cyberattack" alone is therefore too coarse for a defensible insurance claim or a scientifically meaningful severity distribution.

Second finding: contract design determines the recoverable share

Property grants, digital-data provisions, ensuing-loss language, cyber response grants, business-income triggers, waiting periods, exclusions, and limits can transform the recoverable proportion of the same economic event. The synthetic USD 115 million scenario produces USD 40 million or USD 50 million of residual loss under two stated contractual outcomes. These values arise from assumptions, not an empirical generalization. They demonstrate why enterprise coverage inventories should include exact contract terms and not merely nominal aggregate limits.

A coverage-gap review should use a consistent event narrative across property, cyber, machinery, and liability programs. It should identify physical-damage exclusions, cyber-to-physical write-backs, data-impairment provisions, time-element triggers, dependent-property extensions, shared limits, and the treatment of overlapping loss. Such a review improves the transparency of risk financing even where insurers maintain different legitimate risk appetites.

Third finding: affirmative clarity and underwriting accumulation are linked

PRA, EIOPA, IAIS, Lloyd's, and the LMA address coverage uncertainty from different institutional positions. Their materials jointly establish that exposure to digitally initiated loss can exist in numerous classes, and that explicit terms and exposure measurement matter. The direction of a specific policy revision may enhance or restrict a company's cover. Increased clarity is therefore measurable as a more explicit contractual allocation, not necessarily as a larger amount of indemnity.

For underwriting, shared technology dependencies and physical-process heterogeneity must both be recognized. A common vendor can correlate digital intrusion, while site-specific engineering determines whether that intrusion creates tangible damage, outage, or only a prevented incident. A scientifically meaningful portfolio model separates those mechanisms and evaluates the wording at each layer of insurance.

Fourth finding: insured economic results require observable claim closure

Insured values, reported loss estimates, case reserves, pleadings, arbitration positions, and cash recoveries describe distinct stages. Studies that infer ultimate coverage from either nominal limits or headline gross losses can overstate or understate insurance effectiveness. An evidence-based settlement analysis should reconcile gross enterprise loss, loss qualified under insuring clauses, contractual reductions, legal disputes, and final paid amounts. It should preserve adjudicative uncertainty rather than prematurely replacing an unresolved claim with a zero or full-limit payment.

Future research priorities

The most consequential empirical project is a multinational anonymized dataset connecting industrial event mechanics, equipment damage, interruption duration, loss amounts, actual policy editions, and ultimate indemnity by coverage class. It should specify inclusion criteria, missing-value policies, and denominators, with clear distinctions among attempted attacks, successful OT compromises, realized physical loss, claims reported, claims accepted, and claims paid.

A second project would examine shared-vendor and shared-service dependencies using detailed industrial process models. It should test how safeguards, recovery time, network architecture, and plant operating conditions alter physical severity given the same digital initiating event. A third would compare the operation of cyber exclusion and ensuing-loss forms through actual decisions in identifiable jurisdictions, including outcomes for damage-dependent business income and injury liability.

These are conventional empirical and legal research directions grounded in the demonstrated evidence gaps. They do not establish a new doctrine, model, or predictive framework by declaration. A strong future contribution would be a verified dataset or reproducible comparative finding that changes the present conclusions.

Conclusion

Enterprise cyber-physical damage and its insurance treatment arise from the interaction of operational technology, physical processes, financial consequences, and distinct insurance contracts. Cyber compromise can cause tangible destruction, protective shutdown, service interruption, data impairment, and third-party harm. These effects are materially different, and the public incident record supports different conclusions for different events. A technically sound case therefore begins with a verified causal chain and documents what physically happened.

The contractual response is determined by the insured interest, property and cyber grants, physical-damage and interruption definitions, relevant exclusions and write-backs, time attachment, retentions, sublimits, aggregates, other-insurance provisions, and the applicable legal system. Published judgments show the importance of actual policy wording. Supervisory initiatives favor more explicit identification of cyber exposure across insurance lines, while actuarial scholarship recognizes the difficulty of estimating correlated, infrequent, and highly heterogeneous losses.

The final analytical conclusion is that reliable enterprise insurance coordination requires a single traceable incident history, a nonduplicative ledger of economic losses, and a contract-specific evaluation of every potential payment. The gap between economic loss and actual indemnity is itself an essential financial risk. Clear wording, documented industrial safeguards, consistent evidence preservation, and pre-loss cross-policy testing improve the quality of its assessment. Global empirical claims about the recoverable share of cyber-physical loss await complete, representative engineering and claims records.

References