RESEARCH DOSSIER
Cyber Reinsurance Clash of Coverages and Claims Recoveries
Abstract. Cyber reinsurance clash arises when a digital incident creates liabilities across multiple insurance classes and these liabilities converge within a reinsurance arrangement. The financial consequences depend on original coverage, the assignment of nonduplicated loss items, treaty scope and occurrence definitions, attachments, limits, expenses, inuring order and the collectibility of recoveries. This monograph examines those mechanisms through comparative documentary legal analysis, supervisory evidence, two incident cases, actuarial literature and reproducible hypothetical calculations. Its central finding is that common technical causation is insufficient to determine the allocation of insured and ceded liabilities. Identical primary indemnity can yield different recoveries under distinct contractual groupings, and the direction of this effect changes with the position of losses relative to retentions and limits. The public evidence establishes important mechanisms but does not identify a representative distribution of actual cross-line recoveries. Numerical examples are conditional calculations, not empirical estimates.
Keywords: Cyber reinsurance; cross-line clash; claims allocation; occurrence aggregation; nonaffirmative cyber; treaty exclusions; excess of loss; follow the settlements; recoverables; systemic cyber risk.
Introduction
A digital incident may interrupt production, destroy information, cause financial losses to customers, create third-party liability and produce investigation or defense costs. Its economic effects arise through technical and commercial relationships. Its insurance effects arise through coverage agreements and applicable law. Its reinsurance effects arise through a further layer of agreements between risk carriers. These relationships define the scientific object of this monograph: the convergence of original insurance liabilities across classes and their transformation into ceded claims recoveries.
The study examines which sources establish the scope of cross-line clash, how loss-item allocation and treaty terms transform insured liabilities into financial obligations, and what evidence supports comparative assessment of recoveries. Its parent Science is Cyber Insurance; it intersects with reinsurance, contract law, claims science and actuarial science. The central method is comparative analysis of publicly accessible documents and arithmetic counterfactuals. Legal conclusions are confined to the identified jurisdiction, wording and decision. Institutional guidance is distinguished from an executed policy. No private claims or treaty data are presumed.
Conceptual and Historical Foundations
Units of analysis
A technical incident is a sequence of adverse developments in digital systems. A loss item is an identifiable economic harm to an affected party over a specified period. An insurance claim asserts entitlement under a contract. Covered indemnity is the amount payable after relevant terms are applied. A ceded loss is a liability recognized under a separate reinsurance agreement. A collection is a cash or equivalent settlement of that liability. These units can form many-to-many relationships.
Cross-line clash is the concentration of liabilities from distinct insurance classes or coverage grants arising through connected circumstances. Multi-insured accumulation is the concentration of losses affecting distinct insured risks. A cyber incident can generate either or both. A technology provider might claim its own restoration costs, face liability to customers, and cause dependent business interruption at separately insured enterprises. The loss experienced by a customer and the provider's obligation to compensate that customer are related but cannot automatically be added as independent economy-wide injuries.
A common technical exposure can be represented by a shared vendor, software component, identity service or communications dependency. The relevant coverage grouping may differ from that dependency grouping. A single vulnerability may be exploited independently at different times by unrelated actors, whereas one failed software update may affect numerous systems simultaneously. The identification of shared exposure is a necessary risk-management operation; the grouping of covered liabilities is a contract-specific operation.
The research by Biener et al. (2015), described in the abstract of their empirical insurability analysis, uses 994 cyber loss cases from an operational-risk database and identifies interconnected losses, scarce data and information asymmetry as challenges. Its observations support scrutiny of dependence and measurement limitations; the dataset is not a linked sample of paid original and ceded claims.
Evolution of coverage
The emergence of standalone cyber coverage did not remove cyber-related exposures under preexisting property, casualty and specialty insurance. Nonaffirmative exposure identifies the possibility of losses associated with cyber circumstances under contracts that did not explicitly resolve the relevant exposure. The question remains one of actual wording and applicable law. It does not follow from the absence of a cyber product code that the insurer bears no cyber-related liability.
Lloyd's (2022) Market Bulletin Y5381 describes the phased clarification of affirmative and excluded cyber coverage and introduces requirements for state-backed cyber-attack exclusions in specified standalone business. A portfolio can contain different generations of contractual protection simultaneously. Renewal changes prospective obligations; earlier claims and occurrences require examination of the earlier wording and applicable periods.
Financial structure
Proportional reinsurance shares defined business on agreed terms. Nonproportional reinsurance covers an eligible loss within a layer above an attachment. Per-risk, per-occurrence and aggregate structures differ in the unit to which thresholds apply. Facultative placements address specifically accepted risks, whereas treaty arrangements encompass defined business. The combination of these structures means identical original indemnity can produce different retained outcomes.
The financial boundary of clash depends upon whose portfolio is examined. A direct insurer may face accumulation across its own business lines. A reinsurer may receive liabilities from multiple cedents. A retrocessionaire may encounter a different grouping under another agreement. The initial incident record should be retained through each transfer, while each contractual boundary requires its own eligibility and aggregation assessment.
Primary Documents and the Scope of Cross-Line Clash
Primary coverage classification
The coverage matrix starts with the actual insuring agreement. First-party cyber may address defined restoration, response, extortion or interruption expenses. Property insurance may depend upon covered damage or an extension for digital harm. Crime insurance can address specified financial theft and fraudulent transfers. Technology professional liability can concern a service provider's duties, while privacy or network security liability can cover claims by affected third parties. These categories describe possible forms of exposure, not universal grants.
For every loss item the analyst must determine the insured person, interest, covered cause, temporal trigger, notice condition, damage or liability requirement, geographic scope, exclusions, retentions, sublimits and expenses. Each potentially responding treaty must then be read for the equivalent ceded conditions. Classification by underwriting department alone cannot establish reinsurance eligibility.
Event, cause and duration
Aggregation clauses select a unifying factor for applying a retention or limit. In Lloyds TSB v. Lloyds Bank Group Insurance (2003), the House of Lords discusses the distinction in Axa Reinsurance v. Field between an event and an originating cause. The judgment illustrates that materially different connecting language can produce different groupings. It does not decide whether any particular digital catastrophe is one event.
Consider a hypothetical vulnerability present for several years, exploited by one attacker in March and another in September. A treaty requiring a single specified incident may group the resulting losses separately; a broadly worded originating-cause clause might support a wider association, subject to causation and contractual interpretation. A duration clause may constrain the grouping further. Its consequences depend on the specified start time, covered losses, election mechanism, treatment of overlap and any geographic or system constraints. It is inappropriate to impose a conventional catastrophe hours clause on an agreement that does not contain one.
The timing of economic consequences also differs from event timing. A brief failure may cause weeks of restoration and later litigation. A loss-occurring contract, a claims-made policy and a risks-attaching treaty could assign related facts to different years. Every grouping must preserve the underlying time series and the reason for selecting a particular treaty period.
Cyber war and differences in conditions
Y5381 sets minimum requirements for the relevant Lloyd's standalone risk codes CY and CZ, subject to agreed exceptions. It addresses war, state-backed operations significantly impairing state functions or security capabilities, treatment of affected external systems, attribution and definition clarity. This is not a universal insurance statute or a retroactive interpretation of older property policies.
The Lloyd's Market Association (2024) cyber treaty exclusion bulletin describes model clauses LMA5629 and LMA5630; the latter includes a conditional difference-in-conditions mechanism relating to qualifying underlying exclusions. Where underlying language fails the stated conditions, the bulletin describes fallback protection boundaries. The published bulletin is the source considered here. The full model forms and executed treaty wording are not available in this study, so specific clause interpretation is beyond the evidence.
A primary insurer may remain obliged to indemnify while a differently worded ceded agreement excludes the same circumstances. Conversely, a qualifying following mechanism may align the relevant exclusions. Contract analysis must distinguish technical attribution, state involvement, the contract's definition, applicable burden of proof and the evidence available at settlement. Neither the model number nor the existence of a government statement independently establishes coverage.
Governing law and temporal mismatch
In Lexington v. Wasa (2009), the English-law reinsurance period retained its distinct significance even though the primary coverage was construed under Pennsylvania law in a manner encompassing damage outside the coverage period. The case concerned property damage rather than cyber insurance; it supplies a documented analogy for temporal mismatch. Applying it to cyber circumstances requires examination of the actual agreements and applicable law, including any express incorporation or settlement language.
The case is especially relevant when intrusions, discovery and subsequent third-party claims occur in different years. A settlement may reflect primary coverage uncertainty over multiple periods. The allocation under the ceded agreement remains a distinct issue.
Mechanisms and Evidence for Claims Allocation
Reconciling the original loss
Claims allocation begins by identifying the economic harm. A restoration invoice, loss of business contribution, liability judgment and defense expense may represent different items. A single invoice presented under two grants remains the same underlying expense. Where an insured's loss is also the basis of another insured's liability, the analyst must trace payments and potential subrogation or contribution rather than assume independent economic damage.
Suppose a USD 6 million expense falls under two indemnity policies. Assume each would cover the full amount separately, but their operative clauses and governing law assign equal shares. Each insurer pays USD 3 million. A loss register that records USD 6 million twice overstates the underlying expense. If the same incident produces separate USD 4 million interruption loss, the economic total in this simplified example is USD 10 million; its insurance allocation requires separate analysis. Removing duplicative amounts and aggregating genuinely distinct covered items are different operations.
An interruption claim should preserve the financial baseline, mitigation and saved expenses, covered period and valuation assumptions. A technical outage timestamp is evidence of system behavior, not necessarily of the contractual interruption period. Defense, adjustment and extra-contractual amounts should remain separately classified because their treatment can change the amount eligible for a treaty.
Salvage and third-party recoveries may affect ultimate retained losses and the amount due from reinsurers depending on contract terms. A ledger should retain the initial estimate, subsequent revisions, recoveries and timing. Subtracting the same receipt from two independent accounts without an allocation rule can understate total ultimate net loss.
Follow the settlements
A follow-the-settlements provision can restrict unnecessary relitigation of underlying claims, subject to its own conditions and governing law. In Utica Mutual v. Fireman's Fund (2020), the Second Circuit discussed good faith, reasonableness and the terms of applicable policies, and reversed an award in a dispute about underlying aggregate limits. The noncyber case illustrates that settlement deference has contractual boundaries.
Cyber settlements may occur before technical investigation and litigation establish every fact. An allocation should record contemporaneous evidence, alternative coverage interpretations and the basis of compromise. The financial impact upon a treaty is relevant for conflict identification, but favorable reinsurance consequences do not by themselves prove that an otherwise supported allocation was improper.
Claims cooperation, control and notification clauses require examination of their precise text. A practical claim record includes original policy and endorsement versions, the relevant treaty, dates of notice, authority for settlement, the incident chronology, an itemized allocation and a reconciliation of outstanding and paid amounts. Sensitive security evidence should be handled under applicable confidentiality and transfer constraints.
NotPetya and property insurance
The Merck v. ACE appellate opinion (2023) concerned the 2017 NotPetya incident and an all-risk property insurance program with USD 1.75 billion of limits in three layers above a USD 150 million deductible. The decision reproduces policy language expressly addressing categories of data or computer impairment. The New Jersey Appellate Division affirmed the finding that the hostile/warlike exclusion did not apply on the circumstances and wording at issue.
The New Jersey dismissal order (2024) records the later stipulated dismissal with prejudice of the appeals. The order contains no Supreme Court merits ruling. The stated program limits are neither an established settlement value nor a measure of ceded recovery. The case establishes the existence of a potentially material property-coverage interface, not a verified amount paid under cyber reinsurance.
The scientific significance is that a risk register restricted to affirmative cyber product codes may miss digital exposure held under a property portfolio. What the insurer must pay and what its reinsurer must reimburse remain matters of distinct contracts.
CrowdStrike and shared technology failure
CrowdStrike's (2024) Channel File 291 announcement identifies a July 19 content configuration update for Windows sensors as the cause of a widespread outage. Microsoft's contemporaneous outage impact estimate placed affected Windows devices at approximately 8.5 million. These observations supply direct evidence of a common technological dependency and a nonmalicious failure mechanism.
A device count is not an insured count: firms operate different numbers of machines, may have several policies or none, and experience different interruption periods and financial consequences. Downstream firms can suffer disruption without directly operating the affected software. The cases demonstrate two different routes to broadly distributed cyber-related loss. Neither provides a representative estimate of insurance claims or treaty collections.
Comparative Measurement of Reinsurance Recovery
Ceded loss and layer calculation
For each distinct primary indemnity x_i, let y_i represent the treaty-eligible amount after applicable scope, expenses, prior cessions and temporal conditions. The eligible loss for event e is the sum of its assigned y_i. For an illustrative full-share excess-of-loss layer with attachment A and limit M, the contractual amount is R(L) = min(M, max(0, L - A)). A reinsurer's agreed participation applies to the relevant layer according to its terms.
Separate occurrence recoveries can be constrained by a remaining annual aggregate and reinstatement provisions. The examples assume sufficient aggregate capacity unless explicitly stated. They use conventional layer arithmetic, not a new actuarial theory.
Fixed losses and alternative groupings
Assume distinct gross indemnities of USD 12 million in standalone cyber, USD 10 million in property and USD 8 million in technology liability. These amounts are already after original deductibles, exclude adjustments and total USD 30 million. A hypothetical multiline treaty provides USD 20 million excess of USD 10 million per occurrence.
| Assumed treatment | Eligible amounts (USD million) | Recovery (USD million) | Gross less recovery (USD million) |
|---|---|---|---|
| One event, all lines eligible | 30 | 20 | 10 |
| Three events, all lines eligible | 12, 10, 8 | 2 | 28 |
| One event, property excluded | 20 | 10 | 20 |
| Cyber only | 12 | 2 | 28 |
One occurrence produces a full layer payment of 20; the separate grouping produces only 2. The effect reverses in a different region of the loss distribution. Two eligible occurrences of USD 28 million each produce USD 18 million each, or USD 36 million combined assuming sufficient annual capacity. Grouping the same USD 56 million produces USD 20 million. If only USD 20 million of annual aggregate remains, both treatments are capped at USD 20 million. Consequently, the direction of an aggregation effect depends upon thresholds, limits and remaining aggregate. A preference for aggregation cannot be inferred from gross losses alone.
Inuring and expenses
Assume gross eligible loss of USD 30 million, a 40% quota share, and the occurrence layer just described. When the quota share inures to the excess layer, it pays 12, the excess layer applies to 18 and pays 8, and total recovery equals 20. If a different pair of agreements explicitly provides the reverse order, the excess layer pays 20, the quota share applies to the remaining 10 and pays 4, and total recovery equals 24. Naively applying both to gross produces 32, inconsistent with the stated nonduplicated indemnity example.
If USD 4 million of additional adjustment expense is eligible inside an already exhausted occurrence layer, marginal recovery is zero. If an alternative agreement specifically reimburses 50% of those expenses outside the layer, marginal recovery is 2. These differences are contractual rather than statistical. Their treatment requires explicit expense records and coordination terms.
Dependence and severity
For losses with finite second moments, portfolio loss variance equals the sum of individual variances plus twice the sum of pairwise covariances. Shared vulnerabilities and infrastructure may induce dependence. Zeller and Scherer's (2024) accumulation risk study explains how data collection that fails to preserve common vulnerability information can understate dependence and portfolio risk. The implication for clash is that separate claims systems must preserve a means of identifying common incident exposure.
Assume 100 units each face a one-million-dollar loss with 1% annual probability. Under full dependence, all 100 incur loss together with probability 1%. Under independence, the individual probabilities are unchanged. Expected gross loss in both cases is USD 1 million; the probability of all 100 losses in the independent case is 0.01 to the power of 100, compared with 0.01 under complete dependence. These extreme assumptions illustrate tail sensitivity, not realistic market calibration.
If the dependent loss qualifies as one occurrence, it exhausts the example layer, yielding USD 0.2 million expected annual recovery. If each independent loss is stipulated to be a distinct occurrence, no individual loss attaches. This comparison varies both statistical dependence and legal grouping. The fixed-loss illustration above independently demonstrates the sensitivity to contractual classification.
Market reporting and denominators
The NAIC (2026) cybersecurity insurance market report gives about USD 9.464 billion of direct written premium including alien surplus lines for 2025, against USD 9.144 billion for 2024, with reported growth of 3.50%. Written premium measures the specified market reporting base, not ceded claims or world economic losses.
Cremer et al. (2024), in their interview-based reinsurance study, report research based on 20 interviews with participants in cyber reinsurance and describe capacity requirements and diversification practices. Qualitative interviews illuminate practices but do not establish a representative distribution of actual claims recovery amounts.
A recovery ratio must identify its numerator, denominator, observation unit and maturity. Ceded recoverable divided by eligible incurred losses differs from cash collected divided by amounts billed. Gross paid loss divided by written premium describes yet another financial relationship. Comparing ratios without matched valuation dates, policy populations or treaty years risks false equivalence.
A representative longitudinal investigation would need linked information on incidents, insured entities, coverage sections, policy periods, treaty terms, amounts billed, collections, disputes, expenses and currency dates. Public premium reports and incident descriptions do not provide this joint dataset. No market-wide numerical recovery efficiency is inferred in this work.
Financial, Legal and Institutional Consequences
Ultimate loss, credit and liquidity
Contractual reinsurance recoverables and actual cash collection differ. A cedent paying USD 30 million immediately and receiving USD 20 million six months later funds USD 30 million initially and retains a net nominal cash outflow of USD 10 million afterward. At a hypothetical annual discount rate of 5%, the six-month receipt has a present value of approximately USD 19.52 million. This calculation is a financing example and does not prescribe regulatory accounting.
Counterparty default, disputed scope, collateral access and settlement timing may reduce or defer the economic benefit of reinsurance. Severe cross-line events can coincide with stressed counterparties and disruptions to insurers' own operations. Credit and timing assumptions should therefore be evaluated conditional on loss scenarios rather than treated as necessarily independent risks.
Supervisory evidence
EIOPA's (2022) nonaffirmative cyber supervisory statement addresses governance, coverage review, concentration and accumulation assessment and the evaluation of reinsurance capacity. Its recommendations concern national supervisory authorities and the relevant European insurance framework; they do not establish private treaty obligations.
The PRA's cyber underwriting risk statement identifies nonaffirmative exposure, strategy and appetite, and expertise as supervisory areas. Its current-version page records an update effective December 31, 2024. This study relies on that identification and version information.
The IAIS (2023) GIMAR special-topic announcement describes the assessment of cyber underwriting activities in its 2022 monitoring sample and substantial gaps in measuring nonaffirmative exposure. The reported financial-stability assessment was conditional on the market exposure and sample then considered. It should not be applied indiscriminately to future portfolios.
These documents support the identification and governance of concentration risk and the examination of actual reinsurance applicability. Managing an insurer's own operational resilience is a distinct activity from assessing its liability to policyholders, even though the two can interact in a crisis.
Contractual alignment and international disputes
A contractually excluded exposure may be deliberately retained. The problem is unrecognized divergence between the insurer's gross obligations and its assumed ceded protection. Comparisons should identify all major coverage combinations, relevant wording generations, expense rules, dates and remaining aggregates.
Corporate groups can be exposed to multiple legal systems through local policies, cross-border operations and reinsurance arrangements. Choice of law, contractual forum, evidence access, confidentiality and legally relevant restrictions can influence claims handling. A shared incident record can support consistent factual analysis; it cannot automatically transfer a legal interpretation from one agreement to another.
Policyholder obligations depend upon original coverage and applicable law rather than the cedent's subsequent collection. Capital planning, underwriting strategy and claims operations must account for this separation. Where performance of the original obligation precedes settlement of reinsurance, access to funding can become consequential for the insurer's operations and insured outcomes.
Integrated Scientific Synthesis and Research Agenda
Interactions between contractual and statistical processes
The central synthesis is that common-cause dependency, original coverage, allocation of distinct economic losses, treaty eligibility, occurrence grouping and collection involve different operations. A model calibrated only to reported direct cyber premiums cannot identify them. A complete event description is likewise insufficient to settle a claim. Financial risk assessment requires a documented transformation from economic harm into legal liability and contractual recovery.
Grouping can increase or reduce payment. The fixed-loss scenarios establish this directly through conventional layer mechanics. Scope changes can alter retained loss without changing underlying economic circumstances. Inuring can change the attachment base without changing the gross indemnity. Expense eligibility and collection timing can create further divergence. These mechanisms explain why nominally comparable insurance portfolios and limits can produce materially different net financial outcomes.
Requirements for empirical investigation
A linked loss register should distinguish incident, affected party, economic item, policy, claim, treaty and cash settlement identities. Dependence information should permit representation of shared services and components over time. Policy and treaty versions should be recorded with effective intervals and exact wording. Financial fields need consistent valuation dates and treatment of revisions. Disputed and undisputed amounts should remain distinct.
Future frequency research must choose units and denominators explicitly. Policy-based claim rates require exposed policies or insured-years; incident rates require observable incident populations; recovery studies require eligible losses or billed recoveries and settlement maturity. Observations of still-open claims are censored rather than proof of permanent nonpayment.
Selection is another challenge. Different cedents negotiate different contracts and serve different businesses. Higher recoveries may reflect richer coverage, more severe loss, lower attachment, different claim maturity or different portfolio composition. Identifying a causal effect of a wording change requires control of those factors and a plausible comparison design. The available public record does not support such causal estimation.
A further research program should examine complete model and executed clauses, collect jurisdiction-specific legal outcomes, and obtain anonymized claims-level gross-to-net records. Tests should separately vary loss grouping, treaty scope, expense treatment, prior cessions and aggregate availability. Where probabilities cannot be established, the resulting figures should be reported as sensitivity calculations rather than probabilistic forecasts.
Scope and original analytical contribution
This work is a selected comparative documentary study, not an exhaustive systematic review. Complete privately executed treaties, linked market recoveries and representative claims files were unavailable; the LMA source supports bulletin-level observations. The Biener et al. article is used at abstract level. The official NAIC report and New Jersey dismissal order were examined through extracted text; their image readback was unavailable. No inference is made from unobserved private settlements.
The original contribution is the integrated explanation of primary cross-line liability, assignment of nonduplicated indemnities, legally defined event grouping, financial layer mechanics and separate collection exposure. The numerical scenarios are reproducible and conditional. They apply established methods, do not claim empirical novelty, and introduce no named model, doctrine or theory.
Conclusion
Cyber reinsurance clash arises through the convergence of cyber-related original liabilities across insurance classes within risk-transfer arrangements. Technical dependence identifies potentially correlated losses. Coverage and applicable law determine primary obligations; treaty wording determines ceded eligibility and grouping. Financial terms determine recoverable amounts, and actual settlements establish collection.
The legal cases establish the importance of wording and the distinct boundaries of original and ceded contracts. The institutional documents identify exposure recognition and concentration governance as important supervisory concerns. The incident evidence identifies different technical routes to shared loss. The actuarial literature supports collecting information on common dependencies rather than relying solely on marginal claim experience.
The worked scenarios establish that fixed gross loss can yield different recoveries under alternative grouping, scope and inuring conditions. The direction of the grouping effect changes with the layer and available aggregate. Actual cash funding may remain larger than ultimate retained loss while reinsurance is unpaid.
A complete investigation of market outcomes therefore requires reconciled loss and contract data with explicit valuation dates, exposure denominators and settlement histories. Publicly observable premiums, technical incident counts and court cases cannot supply that distribution by themselves. Scientific analysis can explain the mechanisms while identifying precisely the evidence required for further measurement.
References
Biener, C., Eling, M., & Wirfs, J. H. (2015). Insurability of cyber risk: An empirical analysis. The Geneva Papers on Risk and Insurance—Issues and Practice, 40(1), 131–158. https://ideas.repec.org/a/pal/gpprii/v40y2015i1p131-158.html
Cremer, F., Sheehan, B., Mullins, M., Fortmann, M., Materne, S., & Murphy, F. (2024). Enhancing cyber insurance strategies: Exploring reinsurance and alternative risk transfer approaches. Journal of Cybersecurity, 10(1), tyae027. DOI: 10.1093/cybsec/tyae027. https://academic.oup.com/cybersecurity/article/10/1/tyae027/7920185
CrowdStrike. (2024, August 6). Channel File 291 incident: Root cause analysis is available. https://www.crowdstrike.com/en-us/blog/channel-file-291-rca-available/
European Insurance and Occupational Pensions Authority. (2022). Supervisory statement on management of non-affirmative cyber exposures. EIOPA-BoS-22-414. https://www.eiopa.europa.eu/system/files/2022-09/supervisory_statement_on_management_of_non-affirmative_cyber_exposures.pdf
International Association of Insurance Supervisors. (2023, April 21). IAIS GIMAR special topic edition provides assessment of cyber risks in the insurance sector and financial stability implications [Press release]. https://www.iais.org/2023/04/iais-gimar-special-topic-edition-provides-assessment-of-cyber-risks-in-the-insurance-sector-and-financial-stability-implications/
Lexington Insurance Co. v. AGF Insurance Ltd. and Lexington Insurance Co. v. Wasa International Insurance Co. Ltd., [2009] UKHL 40 (House of Lords, July 30, 2009). https://publications.parliament.uk/pa/ld200809/ldjudgmt/jd090730/lexing-1.htm
Lloyd's. (2022, August 16). State backed cyber-attack exclusions. Market Bulletin Y5381. https://assets.lloyds.com/media/35926dc8-c885-497b-aed8-6d2f87c1415d/Y5381%20Market%20Bulletin%20-%20Cyber-attack%20exclusions.pdf
Lloyd's Market Association. (2024, March 1). Model state-backed cyber war exclusion clauses for cyber treaty reinsurance. Bulletin LMA24-011-CM. https://my.lmalloyds.com/WP/LMA_Bulletins/LMA24-011-CM.aspx
Lloyds TSB General Insurance Holdings Ltd. v. Lloyds Bank Group Insurance Co. Ltd., [2003] UKHL 48 (House of Lords, July 31, 2003). https://publications.parliament.uk/pa/ld200203/ldjudgmt/jd030731/lloyds-1.htm
Merck & Co., Inc. v. ACE American Insurance Co., Nos. A-1879-21 and A-1882-21 (N.J. Super. Ct. App. Div., May 1, 2023). https://law.justia.com/cases/new-jersey/appellate-division-published/2023/a-1879-21.html
Merck & Co., Inc. v. ACE American Insurance Co., Nos. A-62/63, 088274 (N.J. Supreme Court, January 23, 2024; filed January 26, 2024) [Dismissal order]. https://www.njcourts.gov/system/files/court-opinions/2024/a_62_63_22.pdf
National Association of Insurance Commissioners. (2026). Report on the cybersecurity insurance market. https://content.naic.org/sites/default/files/2026-report-cybersecurity-insurance-market.pdf
Prudential Regulation Authority. (2024, November 15). Cyber insurance underwriting risk (SS4/17, November 2024 update). Bank of England. https://www.bankofengland.co.uk/prudential-regulation/publication/2017/cyber-insurance-underwriting-risk-ss
Utica Mutual Insurance Co. v. Fireman's Fund Insurance Co., 957 F.3d 337 (2d Cir. 2020). https://law.justia.com/cases/federal/appellate-courts/ca2/18-828/18-828-2020-04-28.html
Weston, D. (2024, July 20). Helping our customers through the CrowdStrike outage. Microsoft. https://blogs.microsoft.com/blog/2024/07/20/helping-our-customers-through-the-crowdstrike-outage/
Zeller, G., & Scherer, M. (2024). Is accumulation risk in cyber methodically underestimated? European Actuarial Journal, 14, 711–748. DOI: 10.1007/s13385-024-00381-9. https://link.springer.com/article/10.1007/s13385-024-00381-9