Prolepsia forecastYes / no
Will a single cyber event cause insured losses above USD 1 billion in 2027?
As of 15 November 2026, Prolepsia puts the chance of yes at 19%.
19%chance of yes
How Prolepsia reasons
The report Prolepsia wrote with this forecast.
How forecasts are made · Track record
TL;DR
The forecast is 18.7% Probability of Yes and 81.3% Probability of No. A qualifying event is a material risk, but the stronger expectation is that no single 2027-origin incident establishes losses strictly above USD 1 billion under the question’s narrow insurance definition.
Key evidence
- The strongest recent analogy developed downward. CrowdStrike’s initial estimates included Guy Carpenter’s USD 300 million–1 billion range and CyberCube’s USD 400 million–1.5 billion range. Guy Carpenter’s January 2026 chart instead indicates approximately USD 250 million. That is an approximate industry estimate, not a final claims determination, but it weakens the case that CrowdStrike demonstrated a qualifying billion-dollar loss (initial comparison; later chart).
- Historical headlines use broader coverage definitions. NotPetya exceeded USD 3 billion in all-lines insured losses, but approximately 85% during its first two development years was non-affirmative property loss. Most of that headline total therefore does not belong in this question’s reference class (PCS analysis).
- Large insured events remain credible. A study of affirmative US cyber exposure models a USD 9.94 billion single-event loss at a 100-year return period. This supports substantial catastrophe potential, not a direct global frequency estimate at USD 1 billion (Guy Carpenter/Guidewire study).
- Coverage can expand the loss pathway. AIG introduced a cloud-outage endorsement with a two-hour waiting period and no subsequent monetary retention. Its industry-wide significance depends on undisclosed take-up and insured limits (AIG announcement).
Uncertainties
The largest unknown is the frequency of prolonged common-cause incidents affecting many substantially insured firms. Waiting periods, limits, exclusions, restoration speed and the distinction between lost productivity and lost revenue determine how much disruption becomes insured loss.
The historical record is incomplete, so a lack of established qualifying precedents is not proof of zero risk. Resolution also requires defensible event aggregation and developed loss evidence through December 31, 2029. Exactly USD 1 billion, or a preliminary range merely crossing it, does not establish YES; genuinely indeterminate reporting can lead to annulment.
Read the full report
TL;DR
The forecast is 18.7% Probability of Yes and 81.3% Probability of No. NO is the clear central expectation for a single event originating in 2027 to establish strictly more than USD 1 billion in worldwide gross insured losses under affirmative cyber policies. The decisive distinction is between widespread digital disruption and losses that insurance actually covers.
Context
This question covers malicious attacks and accidental technology failures. The incident must originate between January 1 and December 31, 2027, but claims and industry assessments can develop through December 31, 2029. Losses across organizations and countries count when they arise from the same identifiable event. Unrelated attacks cannot simply be combined, even when they share an attacker or a vulnerability.
The coverage boundary is just as important as the event boundary. The measure includes covered business interruption, response, restoration, extortion and liability under affirmative cyber policies. It excludes uninsured damage and payments solely under other insurance lines. Gross insured losses are measured before reinsurance recoveries; those recoveries are not added again. Exactly USD 1 billion does not qualify.
Evidence
The historical backbone supports restraint. Several famous cyber incidents caused enormous disruption, but their headline losses are not clean observations for this question. The public historical record is also incomplete. It supports a judgment that qualifying events are uncommon, not a precise observed annual frequency.
NotPetya illustrates the coverage problem. PCS reported more than USD 3 billion in all-lines insured losses, with two large claims contributing more than 80%. Approximately 85% of the insured loss during its first two development years was non-affirmative property loss. Counting the entire amount would greatly overstate the relevant historical experience (PCS analysis). The Geneva Association likewise separates affirmative and non-affirmative losses in its historical comparisons, rather than treating every cyber-related insurance payment as cyber-policy loss (Geneva Association report).
CrowdStrike is the closest accidental-failure analogy. Guy Carpenter initially estimated USD 300 million–1 billion, while CyberCube estimated USD 400 million–1.5 billion including CrowdStrike’s own insurance tower, or USD 372 million–1.4 billion excluding it. These were preliminary ranges, not evidence that the threshold had been exceeded (initial industry comparison). Guy Carpenter’s January 2026 chart later placed the loss at approximately USD 250 million. The chart supports that magnitude, not an exact final claims total (renewal report). The later CrowdStrike estimate is more informative than the early upper bound.
Policy terms help explain that gap. Guy Carpenter’s initial analysis cited business-interruption waiting periods of roughly four to 12 hours. An outage can attract global attention while affecting relatively few insured firms for long enough to generate covered claims. Even where the waiting period is passed, the relevant loss is not automatically every hour of employee productivity (CrowdStrike analysis).
The October 2025 AWS outage provides another test. CyberCube estimated USD 38 million–581 million and expected losses toward the lower end. Rapid recovery and potential reimbursements limited the expected insurance impact. These were modeled estimates, not final paid claims, but they again show that a large affected population need not produce a billion-dollar insured event (CyberCube assessment).
The November 2026 Microsoft outage supplies a fresh, unresolved example. Kovrr’s preliminary estimate was USD 700 million within a USD 300 million–1.4 billion range. Marsh Re expected minimal insurance-market impact, while CyberCube also expected limited claims despite disruption across more than 75 countries. Uneven disruption and the difference between productivity loss and revenue interruption explain part of the disagreement (Kovrr reporting; Marsh Re assessment; CyberCube assessment). This incident is outside the qualifying year. Its value is as evidence about insurance conversion, not as a candidate for resolution.
Catastrophe models provide the counterweight to this restrained historical picture. CyberCube’s counterfactual research associates a USD 830 million SolarWinds scenario with a five-year occurrence return period, and a distinct USD 803 million WannaCry baseline with the same return period. These are modeled US standalone-cyber results, not observed global frequencies. They place the target within a meaningful insured-loss tail without directly answering the worldwide 2027 question (CyberCube study).
A separate study of affirmative US cyber exposure models occurrence losses of approximately USD 6.8 billion at a 50-year return period and USD 9.94 billion at a 100-year return period. It explicitly distinguishes the largest single-event loss from total annual aggregate loss. That distinction matches this question. The results establish substantial loss potential, but extrapolating a remote tail down to USD 1 billion does not produce a reliable frequency estimate (Guy Carpenter/Guidewire study).
Insured exposure is expanding, although ordinary claims indicators are mixed. NAIC reports US direct written cyber premiums, including alien surplus lines, of approximately USD 9.14 billion in 2024 and USD 9.46 billion in 2025. Domestic policy counts rose from 4,368,614 to 5,843,175, while 2025 claims reached 56,323, up about 20%. More policies create more potential insured victims, but policy growth is not a proportional increase in common-cause catastrophe frequency (NAIC report).
Chubb’s large-account history also resists a simple story of uniformly rising risk. For US insureds with revenue of at least USD 1 billion, the reported series is as follows (Chubb claims report):
| Year | Claims per 100 policies | Mean claim severity, USD |
|---|---|---|
| 2020 | 26.16 | 703,171 |
| 2021 | 22.30 | 557,484 |
| 2022 | 8.91 | 2,639,833 |
| 2023 | 9.40 | 1,421,374 |
| 2024 | 15.16 | 2,216,645 |
| 2025 | 9.95 | 4,428,494 |
The latest year shows higher mean severity and lower frequency. This is one insurer’s portfolio, not a global catastrophe census, and it does not establish a proportional change in systemic-event risk (Chubb claims report).
Recent technical incidents show credible aggregation pathways. LiteLLM demonstrated compromise of a legitimate software-release workflow, while Qubole showed how standing vendor access can reach multiple customers. Microsoft’s ARTEX investigation documented cross-tenant attack attempts, but a high-volume password-spray subset produced no successful authentication. These findings support the possibility of rapid propagation without establishing large covered losses (LiteLLM investigation; Qubole investigation; Microsoft investigation).
Defense and coverage are changing too. Wiz reported sharp reductions in exposed PHP2Shell systems after disclosure, showing that containment can interrupt a potential loss chain. In the other direction, AIG introduced cloud-outage coverage with a two-hour waiting period and no subsequent monetary retention. Faster containment reduces insured severity; broader coverage can increase it. AIG’s undisclosed take-up and aggregate limits prevent a firm market-wide conclusion (remediation evidence; AIG announcement).
What's non-obvious
The obvious narrative focuses on how many organizations depend on the same technology. The less obvious issue is how many suffer covered financial loss from the same originating incident. Reach, compromise, business interruption, insurance coverage and accepted event aggregation are separate hurdles. A dramatic economic-loss headline can therefore be weak evidence for this particular threshold.
The reverse also matters. Modest premium growth does not prove static exposure: prices, policy counts, limits and coverage wording can move differently. My reading is that the strongest upside risk is not simply more attacks. It is a persistent common-cause incident that reaches revenue-critical systems at many insured firms while coverage terms allow their losses to accumulate.
Uncertainties
The largest gap is a complete, consistently defined ledger of developed worldwide affirmative-cyber event losses. The available history is selected and sometimes mixes policy lines or geographies. It cannot support treating the absence of a verified qualifying precedent as a verified zero-event history. Event-level developed claims assessments would most improve this part of the analysis.
The main forward uncertainty is the frequency of sufficiently severe common-cause events. Better information on shared-provider dependencies, insured limits, waiting periods and coverage take-up would help distinguish a broad outage from a billion-dollar insured catastrophe. Technical demonstrations and attempted intrusions do not close that gap.
Finally, claims development can change the outcome near the threshold. The assessment deadline favors developed industry-wide claims evidence over early projections. Without such an assessment, the criteria require independent supporting estimates above the threshold, not merely ranges crossing it. Liability disputes can outlast the deadline, and genuinely insufficient evidence can lead to annulment. The forecast applies no separate annulment discount.
Forecast history
| Date | Forecast |
|---|---|
| 19% chance of yes |
Sources Prolepsia read
- PCS analysiswww.verisk.com
- https://www.genevaassociation.org/sites/default/files/2024-12/cyber_ils_report_1213.pdfwww.genevaassociation.org
- initial industry comparisonwww.guycarp.com
- Headingwww.actuarialpost.co.uk
- CyberCube Shares Insurance Loss Estimate for AWS “Amazonk” Outagewww.cybcube.com
- Microsoft outage could generate $700 million in insured losses: Kovrr | The Insurerwww.theinsurer.com
- CyberCube studyinsights.cybcube.com
- REPORT ON THE CYBERSECURITYcontent.naic.org
- Chubb claims reportwww.chubb.com
- LiteLLM Compromised, Steals Credentials | Wiz Blogwww.wiz.io
The question
This question asks whether at least one cyber event occurring between January 1, 2027, and December 31, 2027, inclusive, will cause more than USD 1 billion in aggregate insured losses worldwide. The threshold applies to the total insured losses attributable to a single event across all affected policyholders, insurers, and jurisdictions, rather than the losses sustained by any individual company or insurer. A qualifying cyber event may involve a ransomware attack, widespread data breach, supply-chain compromise, malicious exploitation of a software vulnerability, distributed denial-of-service attack, cloud infrastructure failure, or accidental software malfunction that causes significant digital disruption. Both malicious and non-malicious events qualify, provided the resulting losses are covered by insurance. The relevant measure is the global aggregate gross insured loss, before reinsurance recoveries, under affirmative cyber insurance policies, including covered business interruption, incident response, data restoration, cyber extortion, and third-party liability claims. The question concerns insured losses rather than total economic damages, which may be substantially higher. As background, the July 2024 CrowdStrike software update failure demonstrated the potential for a single technology incident to generate substantial insured losses across multiple industries and countries. Guy Carpenter estimated insured losses from that incident at USD 300 million to USD 1 billion, while CyberCube estimated USD 400 million to USD 1.5 billion, illustrating both the potential scale of systemic cyber events and the uncertainty surrounding industry loss estimates. The Geneva Association has also documented the growing significance of cyber catastrophe risk and noted that Verisk's Property Claim Services classifies cyber events exceeding USD 250 million in insured losses as cyber catastrophes. As of November 15, 2026, the question is forward-looking and concerns events originating during calendar year 2027, with additional time allowed for insurance claims and industry loss estimates to develop. Background sources include Guy Carpenter's August 2024 analysis (https://www.guycarp.com/content/guycarp-rebrand/en/insights/2024/07/global-outage-with-widespread-impact.html) and the Geneva Association's December 2024 report (https://www.genevaassociation.org/sites/default/files/2024-12/cyber_ils_report_1213.pdf).
How it resolves
The question resolves YES if at least one qualifying cyber event whose initial triggering incident occurs between January 1, 2027, and December 31, 2027, inclusive, is determined to have generated more than USD 1,000,000,000 in aggregate worldwide gross insured losses under affirmative cyber insurance policies. The threshold must be strictly exceeded; an estimated loss of exactly USD 1 billion does not qualify. The question resolves NO if no qualifying event is found to have exceeded this threshold by the final assessment date of December 31, 2029, based on the available evidence. Resolution will rely primarily on event-specific industry insured-loss estimates or assessments published by established insurance loss reporting organizations, including Verisk Property Claim Services (PCS), where available. Supporting sources may include catastrophe modeling and insurance analytics firms such as CyberCube, Guy Carpenter, Aon, and other recognized insurance industry research organizations. Where an authoritative industry loss estimate is available, the most recent credible estimate published by the assessment deadline will be preferred over earlier preliminary estimates. If multiple credible estimates disagree, priority will be given to a reported industry-wide loss assessment based on developed insurance claims data rather than an early catastrophe model projection. If no such assessment is available, at least two independent, credible industry sources must provide point estimates exceeding USD 1 billion, or provide estimated ranges whose lower bounds strictly exceed USD 1 billion, for the event to qualify. An estimated range that crosses the threshold, without sufficiently strong additional evidence supporting an outcome above USD 1 billion, will not establish a YES resolution. Loss estimates must represent aggregate insured losses attributable to the same identifiable event, not the sum of unrelated incidents occurring during the year. Losses may develop or be paid after 2027, provided they arise from an event originating in 2027. The assessment will use information published through December 31, 2029, without reopening the outcome for subsequent revisions. If reporting is insufficient to establish whether the threshold was exceeded, and the available evidence does not support a reliable determination either way, the question will be annulled rather than automatically resolving NO. All monetary amounts will be expressed in nominal USD, with non-USD losses converted using the exchange rates employed in the underlying industry assessments or, if unavailable, prevailing exchange rates at the date of the event.
Fine print
A single cyber event includes a common-cause incident affecting multiple organizations, such as a coordinated ransomware campaign, compromise of a widely deployed software product, outage of a major cloud provider, or exploitation of a shared vulnerability. Multiple affected organizations may be aggregated when their losses are directly attributable to the same identifiable originating incident. However, separate attacks by the same threat actor, or unrelated compromises exploiting the same general vulnerability, will not automatically be aggregated unless recognized industry assessments explicitly treat them as one event. For incidents spanning multiple calendar years, the event is assigned to the year in which the initial triggering incident occurred, provided that date can be established with reasonable confidence. A cyberattack beginning in December 2027 and producing significant insured losses in 2028 therefore qualifies, whereas an incident beginning in December 2026 does not qualify merely because substantial claims emerge in 2027. The definition includes accidental technology failures and software errors when they cause cyber-related disruption covered by affirmative cyber insurance. Losses paid solely under property, general liability, directors and officers, or other non-cyber insurance policies are excluded, even if the underlying incident was cyber-related. Reinsurance recoveries are not added to the primary insured losses because doing so would double-count the same underlying claims. Uninsured losses, indirect macroeconomic effects, theoretical maximum losses, and speculative damages are excluded. Claims expenses and settlements are included when counted within the recognized insured-loss assessments. The relevant geographic scope is global, with no restriction on where the incident originates or where affected policyholders are located. The December 31, 2029, assessment deadline provides approximately two years after the end of the event window for claims development and reporting, although substantial uncertainty may remain for complex events involving prolonged litigation or delayed settlements.
More public forecasts
Will at least one US Atlantic hurricane cause more than 20 billion dollars of insured losses in 2027?
27%chance of yes
Prolepsia forecast
Will Swiss Re estimate global insured catastrophe losses above $150 billion for 2027?
50%chance of yes
Prolepsia forecast
Will US severe convective storm insured losses in 2027 exceed $50 billion, as estimated by Gallagher Re?
57%chance of yes
Prolepsia forecast
Will global insured natural catastrophe losses in 2027 exceed $120 billion, as reported by Swiss Re?
73%chance of yes
Prolepsia forecast
Will Los Angeles host the 2028 Summer Olympics with no event moved or cancelled for security reasons?
96%chance of yes
Prolepsia forecast
SCOTUS accepts sports event contract case by December 31, 2026?
<1%chance of yes
Prolepsia forecast