Insuring AI loans: who will pay?

Tracing losses from banks to reinsurers, and the customers linking their risks

Documentary investigation · l0g · 11 November 2026. Sources checked as of this date. Figures in the illustrative examples are explicit assumptions; they describe no identified bank insurance policy. Contract documents for the recent transactions remain outside the public sources reviewed.

A bank can finance a data centre, keep the loan on its books and ask an insurer to take on part of the default risk through loan non-payment insurance. Repayments still flow to the bank; if a covered payment default occurs, another balance sheet must provide the money. That protection can sometimes allow the bank to lend more. The question is how many promises, at the end of the chain, depend on the same revenue. S18 S10

On 11 November, the Financial Times reported that Barclays had recently bought cover, HSBC had explored the product and JPMorgan had considered it before deciding against it. The reporting relies on people familiar with the programmes or discussions. These are three distinct situations, and neither the contracts nor the precise risk carriers have been disclosed. S01

The implications extend beyond those banks. The previous day, the UK banking supervisor finalised a reform of its treatment of large exposures. From 1 January 2028, where eligible protection reduces an exposure, the amount deducted will have to be reassigned to the protection provider. Some reassigned exposures will be exempt from the limits while remaining subject to reporting. The rule captures the problem running through this investigation: risk can leave one line of the exposure table and accumulate in another. S07

Insurance has a real economic purpose. It brings other institutions’ loss-absorbing capacity into the financing structure and can spread losses among institutions able to absorb them. Assessing that diversification, however, means looking beyond the names of buildings and banks. One buyer of computing capacity may support several data centres; one reinsurer may protect several insurers; an insurer may also hold bonds issued by companies whose debts it guarantees. The number of contracts gives only a partial picture of the number of independent risks. Supervisory principles specifically call for examination of these interactions between business activities, assets and liabilities. S10

The market predates November’s announcements

Insurance for data centre financing already has a history. In July 2023, The Hartford explained how non-payment cover could allow a bank to increase its commitment to a project. The insurer also described particular caution towards some facilities hosting AI. Although a commercial publication, the document establishes that both the product and this distinction predate the current announcements. S18

A case published by Marsh and Oliver Wyman in June 2024 provides a more concrete example. It described a US bank’s $300 million senior secured term loan in the data centre sector, supported by non-payment insurance. The report discusses the arrangement’s prudential treatment while keeping the bank and insurer anonymous. The $300 million refers to the loan; the fraction of risk actually transferred is undisclosed. The case documents the mechanism’s use according to the report’s authors, while leaving its full financial structure out of view. S06

What makes November’s development significant is its use during a rapid expansion of AI lending. In reporting published on 28 October and carried elsewhere the following day, Bloomberg said Barclays was seeking protection on more than $10 billion of debt associated with this infrastructure. The discussions were still exploratory. That figure does not establish how much is now insured: no public reconciliation connects the newly reported purchase with the earlier report. S02 S01

Another Bloomberg report, on 30 October, described HSBC’s discussions with insurers. It separately mentioned a significant risk transfer (SRT) transaction referencing more than $1 billion of data centre loans. That billion refers to the SRT’s reference portfolio, mentioned separately from the insurance discussions. SRT here describes a structured transaction transferring part of a portfolio’s risk; applying the label to every loan insurance policy would blur a material distinction. S03 S12

HSBC’s own public comments provide useful context. On the 27 October analyst call, chief financial officer Pam Kaur described data centre lending as small and a limited contributor to the group’s growth. Managing concentrations can be ordinary lending discipline. The reported initiatives alone offer insufficient grounds to infer particular concern about the borrowers involved. S04

Three contracts, three different questions

The mechanism starts with the data centre’s commercial contract. A customer reserves capacity, space or computing services, with payments expected to contribute to project revenue. Then comes the loan agreement: the operator or a dedicated vehicle borrows to build, equip or refinance the facility. Finally, the bank may buy a policy covering a defined loss from non-payment. Market publications distinguish this bank protection from insurance on trade invoices. S18 S19 S21

Those distinctions become consequential when a project runs into trouble. A defective server, a delayed power connection or a disputed delivery can affect the customer contract. For the bank, the first question concerns the debt: which entity still owes payment, when is it due, and what recourse is available? The insurance question follows: do that obligation, borrower and event fall within the cover purchased? Financing for the building, financing for the chips and the operator’s receivables may each carry different protections. The Broadcom–Anthropic structure provides another example involving equipment leases and guarantees, with an architecture distinct from loan insurance. S09 S19

Property insurance, which protects a building or equipment against a covered physical event, follows a further set of rules. Announced capacity to insure physical infrastructure therefore tells us little about capacity to insure its borrowings. Similarly, invoice credit insurance and a loan non-payment policy may respond differently to a dispute over performance. The insured obligation defines the relevant scope. S18 S19

Conceptual diagram. The insurers, reinsurers and any retrocessionaires behind the recent transactions have not been publicly identified. The arrows do not map the contracts of Barclays, HSBC or JPMorgan.

More room on the balance sheet can mean several things

For a credit committee, protection can make a loan acceptable when it would otherwise exceed the bank’s internal appetite for a customer or sector. The regulatory calculation has its own criteria. Any relief depends on the applicable regime, the calculation method, the guarantor’s quality and the policy terms. The US case presented by Marsh and Oliver Wyman uses a particular methodology; its conclusions cannot be extended to all US banks. S06 S08 S09

Under the standardised approach, the Basel framework describes a substitution mechanism: the protected share is treated according to the protection provider, while the uncovered share remains linked to the borrower. Among its requirements are protection that is clearly identifiable, legally enforceable and available within an appropriate timeframe. Maturity or currency mismatches and residual risks must be addressed. Each jurisdiction must implement this international standard; it does not approve any particular insurance policy. S08

Capital and funding also need to be distinguished. With unfunded insurance, a bank acquires a promise of indemnification, without receiving an immediate cash reserve equal to the guaranteed amount. Some structured risk transfers, by contrast, have resources provided in advance and held as collateral. The Bank of England highlights the different counterparty exposures and payment timing risks these arrangements create. S12

The UK reform published on 10 November addresses the concentration created by that promise. The PRA notes that the current framework allows an exposure to be reduced without consistently reassigning the corresponding amount to the guarantor or collateral issuer. It considers that material concentrations can consequently accumulate without being fully recognised, monitored or constrained. The new rule takes effect in 2028. Its economic significance is already clear: when one insurer covers loans from several banks, each bank can reduce its borrower exposure while becoming more dependent on the same insurer. S07

For a fixed volume of outstanding loans, well-designed cover can improve how losses are distributed. If banks use all the resulting room to make new loans, the amount financed grows. The overall effect then depends on two developments: sharing an existing risk and creating additional exposures. Public figures do not allow us to calculate the volume of lending enabled by these new policies.

A loss should be counted once

Consider a deliberately simple example. A $1 billion loan suffers a final loss of $400 million after $600 million is recovered from the borrower and collateral. A policy covers 50% of that loss; the insurer has reinsured 60% of its own obligation. The protections are proportional and aligned, with no deductible in this example, and all valid commitments are honoured in full.

The bank receives $200 million from the insurer and retains a $200 million loss. The insurer recovers $120 million from its reinsurer, leaving a net cost of $80 million. The reinsurer bears $120 million. 200 + 80 + 120 = 400. The loan, the direct guarantee and the reinsurance represent successive layers of the same risk. Adding their amounts as though they were separate losses would inflate the exposure artificially.

That equality requires clarity about what is being measured. The insurer’s gross claims payment is $200 million, while its net cost after reinsurance is $80 million. For liquidity, the first figure may matter before the second becomes a reality. If the reinsurer pays late, the insurer must fund the gap. IAIS principles make clear that reinsurance leaves the cedant, the insurer transferring part of the risk, responsible to policyholders for valid claims. S10

Illustrative assumptions, with no calibration to an actual transaction. Premiums, interest, fees, payment delays and defaults by protection sellers are outside this loss-allocation calculation. Recoveries follow the same proportions.

Insurance thus gives the bank an additional claim, with its own collection conditions. Reinsurance gives the insurer another. The ultimate risk bearer may itself transfer a share of its obligations through retrocession. Identifying that final bearer would require the successive contracts, limits, participations and protections at every level. The public information assembled here stops well short of that depth. S10 S11

Buildings can multiply faster than customers

An operating company provides a view of concentration before insurance enters the picture. In its quarterly report for the period ended 30 September 2026, Core Scientific says CoreWeave accounts for approximately 84% of its nine-month revenue and all of its colocation revenue. Colocation involves providing space, electricity and associated services in data centres. Operating multiple sites therefore leaves the company dependent on payments from its main customer. S15

The same report describes the transfer of certain contracts to CW SPV, an indirect CoreWeave subsidiary. CoreWeave remains a primary obligor for the relevant obligations. The new entity adds a legal layer; the filing specifies that the practical ability to enforce recourse would still depend on CoreWeave’s financial condition at the time of default. S15

That concentration can change. Announced contracts with AMD and a neocloud counterparty are due to begin generating revenue in phases from 2027. A neocloud is a specialised cloud capacity provider, often focused on AI computing. As of 30 September, Core Scientific reports 1,120 MW of contracted customer power, of which 485 MW is already billable. The ratio, 43.3%, measures progress towards billing within that scope. It is not a measure of machine utilisation. S15

A documented operating-company example, separate from the bank insurance policies reported by the FT. No connection has been established between these specific financings and the insurers of the banks mentioned. Core Scientific’s interim figures are unaudited.

This example shows why the appropriate unit of analysis extends beyond the borrowing vehicle. The customer supplying the cash, any guarantor and the group to which they belong all matter. Ten or so project companies may depend on a handful of purchasing decisions. Our investigation of Anthropic’s compute commitments follows another overlap between financing and orders, without establishing a link to these bank policies. The Bank of England specifically highlights the AI supply chain’s dependence on investment by a few large groups and the risk of a simultaneous reassessment of their prospects. S12

A reinsurer may meet the same buyer several times

The following scenario describes no actual portfolio. Four projects each borrow $100 million. All depend on the same buyer of computing capacity. A common shock produces, by assumption, a credit loss of $40 million on each loan. With the same protection shares as before, the banks together bear $80 million, the insurers $32 million and their common reinsurer $48 million. The total loss is $160 million.

The four borrowers have separate identities. The commercial shock affecting them is shared. Reinsurance can diversify exposure by spreading it among strong institutions with other risks. It can also bring several paths back to the same balance sheet. Multiple insurers at the front of the chain leave that question unresolved.

Deterministic illustration: four $100 million loans, each suffering a $40 million loss after recoveries. No default probability, crisis frequency or actual exposure is estimated.

The protection provider’s assets also deserve scrutiny. An insurer might hold securities issued by a major technology customer while covering loans repaid from that customer’s orders. In this scenario, a common deterioration would affect both its investments and its insurance obligations. International principles call for consideration of these interactions between assets and liabilities, including concentrations by sector and group. S10

The public record does not allow that combined exposure to be attributed to any particular insurer in Barclays’ reported programme or HSBC’s discussions. It does establish what needs to be checked: insured exposures, risks assumed through reinsurance, securities held and the corporate groups generating the payments. A useful map must preserve gross and net amounts, dates and ownership links. It must also identify the gaps in the available information.

A customer dispute can take several paths

A default sometimes begins far from the bank. A customer disputes late delivery of capacity, insufficient availability or performance under the contract. Depending on its terms, the customer may seek a remedy or a price reduction, or exercise an exit right. The project’s expected revenue then changes, potentially affecting debt repayment. Operators’ disclosures show why delivery, commissioning and contractual recourse matter as much as the announced value of orders. S15

The insurance needs to be examined at its own level. The loan may remain payable despite the commercial dispute; a challenge may also concern the debt’s validity or the lender’s conduct. The policy determines how these situations are treated. Inferring that every customer dispute is automatically excluded from loan cover, based on the wording of an invoice insurance policy, would apply a potentially different rule to the wrong contract. S09 S19

For policies governed by English law, the practice described by law firm Sullivan calls for particular attention to notification duties, insurer consent to amendments of the loan and the treatment of accelerated repayment. A bank may have the right to demand its entire claim immediately while insurance payments remain tied to the original schedule, depending on the policy. The relevant gap is between the date when the lender needs cash and the date when the protection must pay. S19

The word “sanctions” requires the same precision. In its analysis of clause LSW1865, BPL distinguishes a borrower’s non-payment caused by sanctions from a legal prohibition on the insurer performing certain acts. The former may be covered; the latter may require further steps or a licence before payment. This broker commentary on a specific clause establishes neither Barclays’ policy wording nor the outcome of a future claim. S17

A framework for analysis, without an opinion on any particular policy. The documents assembled do not disclose the recent transactions’ waiting periods, exclusions, consent requirements or recourse provisions.

Liquidity risk therefore warrants separate attention. An acknowledged insurance claim may be paid after other obligations fall due. Concurrent claims increase both the money that must be advanced and the work needed to verify them. Reinsurance adds another collection timetable. The UK supervisor requires insurers to monitor concentrations in reinsurance recoverables, capital resources and protections such as collateral or funds withheld. S10 S11

Useful diversification, measured where it matters

The strongest argument for insurance is sound: specialised capital can bear credit risk that some banks want to limit. Insurers select borrowers, impose conditions, collect a premium and arrange protection of their own. Reinsurance can broaden geographical and financial diversification. The IAIS describes these benefits while requiring assessment of counterparties and scenarios involving multiple claims. S10

The Willis survey published on 26 May 2026 illustrates how capacity and selectivity can coexist. Conducted in January among 76 credit and political risk insurers, it describes a market with ample capacity overall, while reporting that several participants are approaching or have reached their data centre exposure limits. This broker finding reflects the responses collected; it provides neither a worldwide total for insured AI loans nor their ultimate allocation among reinsurers. S05

Supervisory data also call for perspective. EIOPA’s dashboard published on 29 October presents European insurance solvency as broadly stable. Its quarterly indicators refer to the second quarter of 2026; annual indicators refer to the end of 2025. In July, the IAIS had described stable aggregate solvency and liquidity positions at the end of 2025. These observations weigh against the idea that the sector as a whole is already weakened. They do not test the new AI financing policies or simultaneous losses that a shared customer could cause. S13 S14

As in our investigation of the strength of the insurer’s promise, actual resilience depends on a specific combination: the carrier’s resources, concentrations in its commitments, collateral quality, the strength of its own protection and the time needed to access funds. A well-capitalised insurer with limited exposure to the same sector and appropriate reinsurance can absorb a loss that would have weighed more heavily on a concentrated bank. Cover involving counterparties already dependent on the same cycle offers less diversification than it may appear to.

The missing document is a map of commitments

Further investigation would require the complete chain for each transaction: the borrower and its controlling group, the projects concerned, the customers supplying revenue, guarantees, the bank’s retained share, the insurers, reinsurers and any retrocessions. Limits and maturities must accompany the names. Gross amounts show payment obligations; net amounts help prevent the same loss being counted several times.

A stress test would then apply the same commercial shock to every project dependent on the customer concerned. It would need to vary recoveries on buildings and equipment, the time required to redeploy capacity, the maturities of protection and settlement dates. A solvent promise that pays late can create a funding need; a guarantee expiring before the loan leaves an uncovered interval. Supervisory frameworks already recognise these dimensions. S08 S10 S11

Public data currently leave this map incomplete for the recently reported policies. That documentary limitation must be kept distinct from the controls exercised within the institutions and by their supervisors. This investigation neither estimates future losses nor establishes an insurer failure.

The issue it identifies is more precise. Banks are looking for partners to keep financing highly capital-intensive infrastructure. Insurance can help, provided that the new risk carriers bring effective loss-absorbing capacity and diversification. The arrangement’s resilience is tested where the contracts meet: at the customers paying the data centres, then at the institutions able to settle several claims at once. That is where the next coverage announcements will need to be brought together.

Method and limitations

This investigation compares dated reporting, corporate disclosures and supervisory documents available as of 11 November 2026. Source notes identify the passages used and access limitations. Statements by insurers, brokers and banks are attributed to their authors. Basel, PRA and IAIS frameworks explain the mechanisms within different scopes and effective dates.

Both simulations are deterministic and use US dollars. In the first: final loss = 1,000 − 600 = $400m; direct insurance payment = 400 × 50% = $200m; reinsurance = 200 × 60% = $120m; insurer’s net cost = 200 − 120 = $80m. In the second, each project loses $40m: bank $20m, insurer $8m, reinsurer $12m. Four projects produce 80 + 32 + 48 = $160m. The chosen shares describe no actual policy and estimate no statistical correlation.

The billable-power ratio is calculated separately: 485 / 1,120 × 100 = 43.3%, rounded to one decimal place. Revenue concentration covers the nine months ended 30 September 2026. The insurers’ identities, their onward transfers and the ultimate accumulation of risks behind the new policies remain unknown in the documents reviewed. No future loss or individual weakness is inferred from those gaps.

Sources and references

1. S01 · Financial Times · Published: 2026-11-11 · Accessed: 2026-11-11 Banks turn to credit insurance to stretch AI lending limits
   Passages and scope

   Opening passages: reported purchase of protection by Barclays, discussions at HSBC, and a review followed by a decision not to proceed at JPMorgan. Reporting relies on people familiar with the programmes or discussions; contracts, risk carriers and amounts actually insured are not disclosed. The full FT article was not obtained; publicly indexed passages and The Banker’s summary provide partial access. The Banker belongs to the same group as the FT, so its coverage is not independent confirmation. Public summary in The Banker.

2. S02 · Bloomberg; authorised republication by Insurance Journal · Original: 2026-10-28; republication: 2026-10-29 · Accessed: 2026-11-11 Barclays Turns to Insurers to Hedge Billions of Data Center Debt
   Passages and scope

   Opening passages and final paragraph: a search for protection on more than $10bn of data-centre debt, with discussions still exploratory and size and timing unresolved. The figure describes the contemplated debt portfolio, not an indemnity or a confirmed insurance purchase. The original report and its republication are a single journalistic source.

3. S03 · Bloomberg; authorised republication by Insurance Journal · Published: 2026-10-30 · Accessed: 2026-11-11 HSBC Seeks to Hedge Data Center Debt With Credit Risk Insurance
   Passages and scope

   Passages on insurance discussions and HSBC’s response; a separate paragraph describes an SRT referencing more than $1bn of loans. No completed insurance purchase is established. The billion-dollar figure refers to the SRT’s reference portfolio, not a policy amount; any overlap between the loans concerned remains unknown.

4. S04 · HSBC · Call held: 2026-10-27 · Accessed: 2026-11-11 Q3 Results Analyst and Investor Call: edited transcript
   Passages and scope

   Page 1 dates the call to 27 October 2026; page 13 contains Pam Kaur’s comments describing data-centre lending as small and not a material growth driver. This is a management statement, without figures in that passage or confirmation of an insurance policy. The filename does not date the call; the initial publication date of the file is not confirmed.

5. S05 · Willis / WTW · Published: 2026-05-26 · Accessed: 2026-11-11 Credit and Political Risk Insurance Capacity Survey and Market Update 2026
   Passages and scope

   Printed pages 5, 7–8 and 10. January 2026 survey of 76 credit and political-risk insurers; capacity as of February 2026 and data-centre premium experience in 2025. Several respondents are nearing or have reached their data-centre exposure limits. This is a broker panel, with Lloyd’s and company-market overlaps addressed; capacity is indicative per transaction, not a worldwide pool. The exact denominator of the data-centre premium subquestion is not specified. PDF report.

6. S06 · Marsh / Oliver Wyman · Published: 2024-06 · Accessed: 2026-11-11 Expanding the portfolio management toolkit
   Passages and scope

   Appendix B, case study 1: a $300m senior secured loan in the data-centre sector accompanied by non-payment insurance. The intermediaries present an anonymous case; the bank, insurer and transaction date are not identified. The $300m is the loan size; the insured share is not disclosed. The case-study table was read in the public index of the primary document; direct access to the PDF was not obtained. No independent confirmation of the transaction is claimed.

7. S07 · PRA / Bank of England · Published: 2026-11-10; effective: 2028-01-01 · Accessed: 2026-11-11 PS20/26: Amendments to the Large Exposures Framework, Part 2
   Passages and scope

   Paragraphs 2.13, 5.5–5.16 and 8.3: mandatory substitution, exposures to protection providers and implementation date. The targeted exemptions concern large-exposure limits; indirect exposures must be reported even when exempt from those limits. This is final UK policy, not yet effective on 11 November 2026; it neither measures AI-loan insurance nor establishes a breach.

8. S08 · Basel Committee / BIS · Version dated: 2020-11-26; effective since: 2023-01-01 · Accessed: 2026-11-11 CRE22: Standardised approach: credit risk mitigation
   Passages and scope

   Paragraphs CRE22.6–22.14 and 22.70–22.84: residual risks, enforceability, maturities, eligible guarantees, substitution and proportional coverage. This version is marked “Current” and has been effective since 1 January 2023; the future November 2028 version is separate. It is an international standard for the standardised approach, applied through each jurisdiction’s rules; it does not approve any individual policy or establish identical treatment under advanced internal-model approaches.

9. S09 · PRA / Bank of England · Published: 2019-07; effective: 2019-09-13 · Accessed: 2026-11-11 SS17/13: Credit risk mitigation, July 2019
   Passages and scope

   Paragraphs 7.2–7.11, printed pages 3–5: eligible guarantees, applicable law, scope of coverage, the protection buyer’s obligations and residual risks. The PRA’s version page identifies this text as current; the January 2026 text takes effect on 1 January 2027. These expectations apply to banks within the stated scope, not to insurers’ solvency calculations. Versions and effective dates.

10. S10 · IAIS · Document version: 2024-12 · Accessed: 2026-11-11 Insurance Core Principles and ComFrame, December 2024
    Passages and scope

    ICP13.0.2–13.0.4, 13.1.8 and 13.5.2–13.5.5: diversification, counterparties, the ceding insurer’s obligation and liquidity. ICP16.1.4–16.1.5 and 16.5.2: stressed correlations and interactions between assets and liabilities. These are international principles, not the terms of an identified contract. Reinsurance does not automatically give the bank a direct claim against the reinsurer. The file is hosted in a June 2025 directory, but the document is the December 2024 version.

11. S11 · PRA / Bank of England · Published: 2024-11-15; effective: 2024-12-31 · Accessed: 2026-11-11 SS20/16: Solvency II: reinsurance, counterparty credit risk, November 2024
    Passages and scope

    Paragraphs 2.1–2.3 and 3.2–3.6, printed pages 2–4: effective risk transfer, reinsurer concentration, stress scenarios, funds withheld, collateral and recoverables relative to capital resources. The stated scope covers UK Solvency II firms, the Society of Lloyd’s and managing agents. These expectations establish neither a lack of supervision nor an absence of concentration. Statement versions.

12. S12 · Bank of England · Published: 2026-07-07 · Accessed: 2026-11-11 Financial Stability Report, July 2026: sections 2.2–2.3 and Box C
    Passages and scope

    Sections 2.2–2.3: dependencies in the AI supply chain and growth in financing. Box C: SRT through securitisation, funded and unfunded protection, counterparty and rollover risks, and regulatory recognition. Not every loan-insurance contract is an SRT. The July findings for the UK do not determine the status of a later transaction or of a foreign bank.

13. S13 · EIOPA · Published: 2026-10-29 · Accessed: 2026-11-11 October 2026 Insurance Risk Dashboard
    Passages and scope

    “Profitability & solvency” and “Liquidity & Funding”, read with the release’s scope. Quarterly indicators cover Q2 2026, annual indicators end-2025 and markets end-September 2026; the release identifies 95 groups and 2,092 solo undertakings. European sector medians are not the ratios of insurers writing AI-loan policies. A solvency ratio measures neither a cash reserve nor a guaranteed share of claims. Release and statistical scope.

14. S14 · IAIS · Published: 2026-07-09 · Accessed: 2026-11-11 Mid-year Global Insurance Market Report 2026: press release
    Passages and scope

    Interim results from the 2026 Global Monitoring Exercise, covering end-2025: stable aggregate solvency, liquidity and profitability, alongside a slight rise in systemic risk scores. This global assessment is not a stress test of AI-loan policies written in 2026. The year-end report is announced for December, after this investigation’s cutoff.

15. S15 · Core Scientific / SEC EDGAR · Filed: 2026-11-05 · Accessed: 2026-11-11 Form 10-Q: quarter ended 30 September 2026
    Passages and scope

    “Concentrations of Revenue and Credit Risk” and the colocation discussion: nine-month revenue, contracted and billed power at 30 September, assignments to CW SPV and recourse against CoreWeave. The issuer’s investor-relations page confirms the 5 November filing date. Interim accounts are unaudited; 485 / 1,120 measures the billable share of contracted power, not machine utilisation. No link to the reported bank insurance policies is established. Filing copy and date on the investor-relations site.

16. S17 · BPL · Anna Morgan · Publication date not displayed · Accessed: 2026-11-11 Sanctions clauses in non-payment policies: a Q&A with BPL Global’s Legal Director
    Passages and scope

    Questions on wording LSW1865: the distinction between a borrower’s non-payment caused by sanctions and acts prohibited for the insurer, including possible applications or licences. This is a broker’s commentary on specific wording; it does not establish its use in Barclays or HSBC policies, or the outcome of a future claim.

17. S18 · The Hartford · Published: 2023-07-06 · Accessed: 2026-11-11 Data Center Financing and Credit Protection
    Passages and scope

    Sections on the role of credit insurance in financing and on data-centre investment risks. This is an insurer’s commercial presentation, reflecting underwriting appetite expressed in 2023. It documents the mechanism’s historical existence, not a named transaction or current appetite across the industry.

18. S19 · Sullivan & Worcester / JD Supra · Marian Boyle · Published: 2022-06-07 · Accessed: 2026-11-11 Non-Payment Insurance Policies: Practical Drafting Tips for Policyholders
    Passages and scope

    Publicly indexed passages on insuring clauses, loan amendments, rescheduling, notification and acceleration. Acceleration of the loan does not automatically accelerate insurance payments, unless the insurer agrees under the policy. Direct HTML access requires JavaScript; no complete direct reading is claimed. This is English-law practice commentary, not universal policy wording or the terms of an identified AI transaction.

19. S21 · European Banking Authority / EBA · Published: 2024-10-03 · Accessed: 2026-11-11 Report on credit insurance: report and press release
    Passages and scope

    Release and publicly indexed primary passages from the report, notably introductory paragraphs 7–9: loan non-payment insurance, dual recourse and the prudential framework. Direct access to the complete PDF returned a 403 error; no full reading is claimed. In 2024, the EBA found insufficient data to recalibrate the treatment and depart from Basel. Its dated observation on insurer defaults does not guarantee future payments; a regulatory loss-given-default estimate is not a probability of default. EBA PDF report.