RESEARCH DOSSIER
Corporate Crime Fidelity and Social Engineering Coverage
Abstract. Corporate crime insurance addresses the financial loss of an enterprise arising from the dishonest conduct of employees, outsiders, and persons who exploit payment systems through deception. The contemporary problem lies at the intersection of traditional fidelity protection, electronic funds transfer, business email compromise, financial crime, payment-services law, and cyber risk. An employee may authorize an outgoing payment that was economically induced by a fraudulent message; a criminal may issue an unauthorized payment order directly to a bank; or an employee may appropriate entrusted corporate assets. These events can produce similar accounting entries while engaging different contractual insuring agreements, different liability rules, and different forensic evidence. This monograph conducts a source-critical, comparative investigation of those distinctions. The documentary corpus comprises insurer coverage descriptions and forms, United States appellate opinions, official U.S. cybercrime complaint statistics, European payment fraud evidence and legislation, United Kingdom reimbursement rules and outcomes, Singapore's responsibility-sharing regime, insurance supervisory standards, and peer-reviewed scholarship. The analysis reconstructs the sequence from fraudulent instruction to transfer, identifies the insured's economic interest and direct loss, distinguishes fraud perpetration from ordinary authorization, and evaluates claim adjustment after bank recovery, third-party restitution, deductible, and sublimit. A reproducible hypothetical calculation illustrates why the same verified transfer may produce sharply different contractual indemnities without implying different underlying criminal conduct. The principal finding is that the legally decisive object is the documented sequence connecting deception, authority, payment execution, and realized loss under the actual wording and governing law. Security controls and payment-regime remedies modify this sequence and its net financial consequences. Aggregated fraud statistics establish the magnitude of reported harms but do not estimate the frequency or expected covered loss of a particular policy portfolio. The resulting synthesis provides a verifiable account of coverage classification, employee–external actor boundaries, digital payment evidence, and uncertainty relevant to underwriting, insurance contract analysis, and claims science. It establishes no new authorial theory or proprietary actuarial model.
Keywords: Corporate crime insurance; fidelity insurance; employee dishonesty; employee theft; social engineering; business email compromise; funds transfer fraud; computer fraud; fraudulent instruction; authorized push payment fraud; digital payments; policy interpretation; direct loss; payment verification; claims evidence; insurance exclusions; credit transfer; cyber insurance; occupational fraud; actuarial uncertainty; international comparative insurance law.
Introduction
An enterprise can lose money because its cashier steals it, because a criminal infiltrates a payment system, because an impostor persuades a controller to send funds, or because an authorized supplier's new bank account is fraudulently substituted. The bank statement records a debit in each case. The criminal mechanism, the insured peril, the payment institution's obligations, and the enterprise's ultimate recoverable loss may nevertheless differ. The financial debit is therefore a necessary piece of proof, while the complete claim requires reconstruction of the event and application of a particular insurance contract.
Corporate crime and fidelity insurance usually operate as first-party protection for specified property or financial losses. The object is the insured organization's loss, even though the operative event is a criminal act. Underwriting and claims practices distinguish employee dishonesty, forgery, computer fraud, funds transfer fraud, and expressly insured social engineering. Travelers' crime coverage architecture presents separate insuring agreements and offers supplemental social engineering applications. This practical separation matters because a fraudulent transfer initiated by a genuine employee may satisfy the social engineering agreement while raising disputes under older computer-fraud language.
The current research question has historical depth. Fidelity insurance arose from the need to protect an organization against dishonesty by entrusted persons, while modern commercial crime programs have adapted to forged instruments, electronic banking, and impersonation across global supplier networks. The relevant historical continuity is the shifting method of appropriating an enterprise's property through its own institutions of trust. Contemporary social engineering extends that vulnerability across human and technical communication channels. Cross and Gillett's review of business email compromise emphasizes the exploitation of trusted relationships and the importance of the human dimension alongside technical compromise.
The subject is financially material. The FBI's 2025 Internet Crime Report identifies 24,768 complaints classified as business email compromise and $3,046,598,558 in reported losses during 2025. The comparable 2024 entries are 21,442 complaints and $2,770,151,146. These are complaint-system statistics. They are neither a census of all corporate crime losses nor insured claims, and no fraction should be identified as indemnifiable without independent coverage information.
The object of this study is the relationship among four records: the occurrence of a dishonest act; the financial institution's execution of a payment; the insured's realized economic loss; and the terms governing an insurer's obligation. The principal questions concern the causal trigger for crime coverage, the position of employees and outsiders in that causal sequence, and the reconstruction of payment loss across jurisdictions. The research examines actual public contract descriptions, appellate decisions and legal provisions, then tests their implications with transparent hypothetical arithmetic. Every public outcome is attributed to its jurisdiction and decision date. It does not infer universal coverage rules from a single appeal.
The documentary method gives priority to operative policy wording as presented in judicial opinions, statutes, regulatory instruments, official statistics and insurer materials. Product marketing is evidence of offered coverage categories, not proof of any insured's policy terms. A court's holding decides the issue before it; a reversal of dismissal does not itself establish a final claims settlement. A reporting agency measures defined transactions or complaints, not an insurance portfolio's latent loss distribution. These distinctions govern the entire analysis.
Scientific Foundations and Comparative History
Corporate crime insurance as a contractual institution
Commercial crime insurance occupies a distinct position in organizational risk finance. Traditional property insurance responds to specified damage or disappearance of tangible property. Liability insurance commonly addresses legal responsibility to third parties. Crime insurance focuses on direct financial or property loss to the insured from designated dishonest acts. The distinctions are functional rather than absolute: policy packages can combine protections, terminology varies among jurisdictions, and a criminal event may simultaneously generate first-party loss, third-party claims, recovery expenses and cyber incident costs.
The market's own documents expose the components of a crime program. Travelers' specimen-form inventory names employee dishonesty, forgery, money and securities, computer fraud and funds transfer fraud, with social engineering addressed in supplemental underwriting materials. Chubb's social engineering endorsement covers specified impersonation schemes, describes a carveback for voluntary parting and discloses its own advertised limits. These are illustrations of real market offerings; their schedules, exclusions and definitions are not interchangeable and the descriptions must never be assembled into a fictitious composite policy.
The conceptual vocabulary is precise. Employee theft or dishonesty concerns the conduct of a person falling within the contract's definition of employee and the requisite dishonest intent or appropriation. Funds transfer fraud may concern an instruction fraudulently purporting to come from the insured and presented to a financial institution. Computer fraud may focus on fraudulent use of a computer that causes a covered transfer. Social engineering fraud addresses deception that induces an authorized person to part with money. The same phishing communication may start more than one of these mechanisms, but coverage depends on the specific insuring agreement, causal language and exclusions.
A firm can experience a theft of its own money, bear an obligation to reimburse an entrusted client, or experience a dispute about whether payment to the wrong account discharged an invoice. Those three legal and economic positions must be separated. The insured must first identify the property, the ownership or responsibility provision and the legal person bearing the loss. If a group company makes payment on behalf of an affiliate, the group structure adds a further question about which policyholder's property was transferred and which entities are insured. A consolidated group cash balance cannot substitute for evidence of individual entity interests.
The structure of an insuring agreement
A policy-level examination begins with the declarations, endorsements and definitions actually incorporated into the contract. It identifies the insuring clause, named insureds, property and funds descriptions, employee definition, territory, covered acts, discovery period, retroactive or prior-loss restrictions, deductible, per-occurrence limit, aggregate where applicable, sublimits, other-insurance provisions and requirements for proofs of loss. A renewal can modify any of these terms. A coverage analysis must record the version effective when the event occurred and, where relevant, the time of discovery or reporting.
Marsh's commercial crime overview describes the named-perils construction commonly used in crime insurance and explains how affirmative social engineering coverage may carry dedicated sublimits. These observations are useful for understanding contemporary program design, while final indemnity always follows the wording issued to the insured. The economic function of a sublimit is especially important: a USD 5 million overall crime limit does not necessarily make USD 5 million available for an impersonation loss.
A claim passes several independent interpretive tests. First, the claimant must show that the loss falls within the insured interest and within the temporal and territorial scope of coverage. Second, the operative criminal conduct must match a covered definition. Third, the conduct must cause the specified form of loss to the required degree. Fourth, exclusions, conditions and limitations must be applied in their proper order. Fifth, actual recoveries and overlapping indemnities must be reconciled. This ordering is an analytical reading sequence rather than a rule of law: an insurer or court may resolve a case at any dispositive step.
A distinctive difficulty arises from words such as direct, resulting directly, unauthorized, fraudulently caused, and voluntarily parted with. A transfer can be formally authorized by an employee yet induced by external deceit. Whether intervening approvals break direct causation is a question of the wording and the governing law. The United States appellate cases discussed below show different outcomes, and their differences supply a controlled comparison rather than a single worldwide definition.
Historical transformation of the financial instrument
In a traditional fidelity episode the employer may identify cash, securities, invoices or an inventory item misappropriated by its employee. Documentary proof may consist of reconciliations, altered checks, payroll records, asset counts, journal entries and the employee's conduct. Digitized payment systems create additional records: an email's underlying headers, identity and device logs, payment-portal events, authentication decisions, bank messages, recipient-account data and recovery attempts. The object remains economic loss, but the proof becomes distributed across institutions.
Business email compromise exploits the organizational division between approval of an economic obligation and execution of the bank transfer. A genuine supplier invoice may coexist with a falsified payment destination. An authentic employee may approve a fictitious acquisition. The attack can succeed without malicious code entering the insured's network. Cross and Gillett (2020), in their analysis of organizational trust abuse, demonstrate why a model of this exposure must account for social relations, enterprise processes and technical communications.
There is a parallel history of regulatory response. The contractual allocation of fraud losses initially depended heavily on bank-customer law and security procedures, while newer payment regimes increasingly require payment service providers to verify recipients or reimburse specified scam victims. The European instant credit transfers regulation and the United Kingdom's APP reimbursement protections illustrate this development. Those regimes shape the insured's net financial position but do not automatically amend a commercial crime policy.
Insurance, bank remedies and cyber coverage
Commercial cyber insurance often addresses incident response, data restoration, business interruption, regulatory liabilities, privacy claims and related cyber-event costs. A fraudulently redirected transfer may engage a cyber product if the actual wording covers the event, but the organizational loss of money must not be presumed to fall within every cyber policy. Some programs use separate crime, cyber and social engineering endorsements; their contribution, other-insurance and excess clauses can affect coordination. The insurer's affirmative grant matters more than the fact that an email or computer was involved.
The NIST Cybersecurity Framework 2.0 provides a useful cross-disciplinary vocabulary of Govern, Identify, Protect, Detect, Respond and Recover. It is a risk-management reference, not an insurance policy, and compliance with it is not a coverage trigger by itself. Its value here is that it distinguishes prevention and detection from post-event response, whereas crime policy analysis concerns the contracted financial consequences after a qualifying event.
Insured Loss Causes and Crime Triggers
The operational chain of a dishonest payment
A fraudulent transfer scheme can be represented as a chronological set of observable events: preparation of an impersonation; delivery of a communication; receipt by an employee or payment service provider; verification or failure to verify; authorization and preparation of a payment order; bank acceptance and execution; receipt by an account controlled by or accessible to the criminal; concealment, conversion or onward transfer; discovery; and attempted recall or restitution. Each step admits different primary evidence. The event chain determines what occurred; legal causation determines which portions of the chain satisfy a policy's requirements.
For example, a vendor-name substitution attack may start with access to genuine correspondence. The impostor sends a plausible invoice showing a new account number. The insured's accounts payable unit updates the master vendor data, obtains legitimate internal approval and transmits a payment order. The bank honors a facially valid instruction. The victim subsequently discovers that the vendor remains unpaid. Every step may be authenticated as a real act by a real employee; the dishonest element is the external alteration of the represented payment destination.
By contrast, in a bank-instruction takeover, the attacker may submit an order purporting to come directly from the insured without an authorized employee's participation. The payment-service institution's obligation to authenticate the order becomes central. In an employee-fidelity case, the actor may be genuinely entrusted with power to initiate transfers but deliberately divert money for personal benefit. In a hybrid case an insider colludes with an external beneficiary. One empirical label such as phishing cannot resolve these alternative factual pathways.
Causation and the significance of wording
The U.S. case law exposes why causal language must be analyzed with the complete transaction record. In Apache Corp. (2016), a caller impersonated a vendor, an email supplied bank details, and the insured made payments after its own verification process. The Fifth Circuit concluded that the use of a computer was only one element in the larger series and did not satisfy the computer-fraud insuring clause. This unpublished decision applied Texas law and the particular policy. It is evidence of a restrictive interpretation on those facts, not a universal requirement that hacking be shown.
The Sixth Circuit reached a different outcome in American Tooling Center (2018). The insured paid a purported vendor account after a fraudulent email exchange. The court reversed summary judgment for Travelers and concluded that the computer fraud caused a direct loss despite the insured's intervening manual steps in making the transfer. The decision is significant because it rejected the inference that a human approval automatically interrupts direct causation. The critical facts were the transfer's relationship to the fraudulent computer use and the wording before the court.
The Second Circuit's Medidata summary order (2018) affirmed coverage for an email-spoofing scheme under the relevant computer-fraud wording. The impersonation involved technical alteration of the apparent origin of an email and subsequent employee action. The opinion is expressly a summary order without precedential effect under its court's local rule; this legal status matters when evaluating its value as comparative authority.
The Eleventh Circuit examined a policy that expressly insured loss from a fraudulent instruction in Principle Solutions (2019). A forged message impersonating a senior manager induced the controller to send more than USD 1.7 million. The court affirmed coverage, construing the particular definition of fraudulent instruction and its direct-loss language. The opinion illustrates the importance of wording that explicitly contemplates an instruction initially received by the insured, rather than only one delivered directly to a bank.
In Ernst & Haas (2022), the Ninth Circuit reversed a dismissal involving an accounts payable clerk deceived into sending money and found the lower court's reasoning about computer fraud and funds transfer fraud erroneous under the facts pleaded. The case was remanded. Its procedural posture permits the conclusion that the claim should proceed under the appellate interpretation; it does not establish that a particular amount was ultimately paid.
These five decisions cannot be reduced to a binary split between computer hacking and human manipulation. They involve different forms, definitions, fact records, appellate postures and states' laws. The comparative result is that an investigator must reconstruct the chain and interpret the policy's causal words before assigning a coverage expectation. A chronology may reveal human steps that do not legally break causation, or technical communications that are insufficiently causal to trigger a narrowly drafted agreement.
Fraud classification and insuring agreement interfaces
The following matrix is a research classification of illustrative fact patterns. Its coverage column identifies the agreement to inspect; it does not predict a binding legal outcome.
| Illustrative event | Key evidence | Main contractual inquiry |
|---|---|---|
| Employee diverts entrusted funds to a personal account | Authority records, bank trace, dishonest purpose, employment relationship | Employee dishonesty and ownership of transferred property |
| Outsider sends a forged payment order directly to the bank | Payment order provenance, signature or token, acceptance log | Funds transfer fraud, bank remedies and unauthorized instruction |
| Outsider persuades controller to issue a genuine wire | Impersonating email, independent approvals, transmission log | Social engineering wording and direct-loss or voluntary-parting clauses |
| Vendor bank details are falsified and internal staff pay the wrong account | Original vendor record, alteration trail, supplier confirmation | Computer fraud, social engineering and resulting debt to supplier |
| Criminal uses malware to issue a transfer without employee approval | Device and bank logs, authenticated sessions, malware evidence | Computer fraud, electronic funds transfer and bank allocation |
| Insider and external actor cooperate | Employee role, communications, benefit flows, collusion evidence | Employee definition, covered act and exclusions for participating principals |
| Client money is held by the insured and misdirected | Client agreement, segregated ledger, custody records | Insured interest and property held for others |
| Payment is blocked before settlement | Bank status, recall confirmation, temporary holds | Whether any covered financial loss has crystallized |
A policy may include overlapping grants, but an insured ordinarily cannot obtain double recovery for the same realized loss. Allocation across grants matters if limits, exclusions, deductibles, and recovery rights differ. It also matters if a criminal act produces separate amounts of direct money loss, forensic expense and consequential business interruption. Such amounts require independent characterization and documentation before they can be adjusted under their corresponding clauses.
Direct financial loss and debt accounting
A victim can pay USD 900,000 to an impostor and remain legally obligated to pay the true supplier for an authentic invoice. The debit to the bank account establishes that the insured transferred USD 900,000. The unpaid supplier liability may demonstrate that the transaction did not discharge the underlying debt. Claim adjustment nevertheless requires evaluation of whether the fraudulent transfer itself represents covered loss, whether the entity owned the money or had legally assumed responsibility for it, and whether recovery or restitution subsequently reduced the amount. Repeated payment of a valid debt must be kept analytically distinct from the first fraudulent outflow to prevent double counting.
Similarly, a large transfer is not automatically a large ultimate loss. A receiving bank may freeze most funds; a beneficiary may repay; an agent may indemnify the enterprise; and a supplier may accept settlement. Claims accounting should identify original principal, recoveries, subrogation, expenses and timing individually. The distinction is particularly important for events that involve multiple legal entities or pooled corporate treasury functions.
Exclusions and the problem of voluntary parting
Voluntary-parting exclusions can create disputes where an employee knowingly instructs a bank to transfer money while misunderstanding the recipient's identity or the business purpose. An explicit social engineering endorsement may modify the result. Chubb's coverage description advertises a carveback to the voluntary-parting exclusion for the product described; the existence of that product demonstrates a market response to the coverage boundary but does not imply that every Chubb form, jurisdiction or policy contains identical language.
Other relevant exclusions may concern trade or investment loss, accounting errors, cyber events, contractual liability, acts of executives, acts by owners with specified interests, prior known circumstances and indirect losses. Whether a particular exclusion applies requires the actual document. No exclusion should be inferred merely because an event included computer access, an authorized user, a foreign beneficiary, or a failure to comply with a suggested control.
Employee and External Fraud Distinction
The actor and the source of authority
Employee dishonesty and external social engineering differ primarily in who performs the dishonest act and the purpose for which authority is exercised. An employee theft claim commonly requires proof that a covered employee acted dishonestly and appropriated property or caused a covered financial loss. An external deception claim commonly requires proof of an impostor's misrepresentation and its effect on a genuine employee's decision. An employee can be a faithful victim of deception while also having breached a payment-control procedure. That breach is an operational fact; the insurance consequences depend on particular contract conditions and governing law.
The distinction cannot be decided by job title alone. Fidelity policies can define employee to include some temporary workers, leased personnel, directors performing specified duties or other agents, while treating owners and independent contractors differently. A company administrator might be authorized to use bank credentials without being an employee as defined by a crime policy. A supplier employee may cooperate with an external criminal but remain outside the insured's employee definition. Underwriting must therefore capture both functional privileges and contractual status.
Attribution becomes especially difficult in a joint enterprise. A parent company may employ a treasury operator who initiates payments from subsidiary accounts. The insured organizations may be listed under a group crime program, but the policy's ownership and insured-person definitions determine whether an act is internal to the relevant insured and which entity suffers the covered loss. The investigator needs intercompany authorization documents, bank mandates, employer-of-record evidence and the operative policy schedule.
Occupational fraud is a separate empirical population
The 2026 ACFE occupational fraud study reports 2,402 investigated cases drawn from 143 countries and territories. Its methodology solicited a single large completed occupational fraud case from responding certified fraud examiners, with the perpetrator reasonably identified. It reports a median case loss of USD 104,000. These cases describe detected internal occupational wrongdoing and are selected by investigator participation and case criteria. They do not estimate the incidence of external BEC in the universe of firms, the ratio of employee theft to social engineering within insured risks, or average crime insurance claim severity.
ACFE's survey is nevertheless valuable for identifying how internal control failures, override opportunities, management power and collusion arise within real investigated cases. The authors' findings do not imply that every internal fraud originates in a uniform organizational profile. The unit of observation is the investigated case, not employee-years at risk. The variable of interest is occupational fraud as defined for the study; insurance payment requires a second, contract-specific classification.
This distinction has an immediate modeling consequence. An insurer cannot combine the ACFE median case loss with FBI BEC complaint frequency to infer an annual fidelity premium. The populations, observation mechanisms, case definitions, financial years and reporting incentives differ. Those datasets describe adjacent processes, not interchangeable estimates for a common statistical parameter.
Trust, communication and authorization
The BEC literature identifies a path through legitimate enterprise relationships. Cross and Gillett (2020) describe how criminals exploit the confidence attached to executives, vendors, clients and professional intermediaries; the risk extends beyond technical account compromise to impersonation and behavioral manipulation. Their scholarly review of trust exploitation supports examining an enterprise's communication procedures together with its technical systems.
An instruction should be analyzed along several dimensions: whether it originates from an authentic account; whether the apparent sender is the actual author; whether the requester has the authority claimed; whether the payment purpose is genuine; whether the destination is controlled by the intended beneficiary; and whether the employee independently verifies the change. Authentication of one dimension cannot logically establish all the others. A genuine executive account can be compromised, a verified phone number can be socially redirected, and a real supplier may have a false bank account substituted into the invoice workflow.
Forensic classification should preserve the difference between an employee authorizing a payment and a criminal authorizing the transaction in a legal sense. The former describes an observed internal action; the latter depends on payment law and agency rules. A manipulated employee may be able to bind the organization under an applicable payment arrangement while the enterprise simultaneously incurs an insured fraud loss. The opposite can also occur: a payment is not legally effective against the customer, leading to bank reimbursement and a reduced or extinguished insurance loss.
Control design as a condition of exposure
A sound treasury-control environment separates initiation, approval, beneficiary maintenance and reconciliation. Unusual payment destination changes can require callbacks to previously established contacts, while independent reconciliations detect deviations after execution. Segregation of duties is most useful when the same person cannot both create a beneficiary and approve a high-value transfer without independent scrutiny. Such controls must be assessed as actual behavior with audit evidence; a written policy alone does not establish that callbacks occurred.
The NIST CSF 2.0 governance categories allow organizations to document assigned responsibility, supplier risk, identity controls, incident response and recovery. The IAIS insurance supervision framework requires insurers themselves to maintain risk-management and internal-control systems. The former can inform insured operational due diligence; the latter governs insurance entities under national implementation. Neither is a direct substitute for an insured's contractual crime-control obligations.
An insurer may request evidence of callback controls, dual authorization, email protections, cybersecurity training, segregation of bank credentials, and formal vendor changes during underwriting. The insurer's actual assessment should distinguish the design of a control, its implementation, evidence of operation over time, and independent testing. The financial incentive is clear: preventing a transfer reduces both loss frequency and claim adjustment effort. It remains scientifically necessary to measure whether a control lowers observed frequency after adjusting for firm size, industry, reporting intensity and other confounders.
Moral hazard and incentives
Economic analysis distinguishes concealment or deliberate insider misconduct from a truthful employee who was deceived. An employee intentionally stealing property may adapt to the monitoring system, while a social engineer adapts to the enterprise's human trust structure. The insured may reduce investment in controls if it expects an insurer to absorb losses, but deductibles, experience-related terms and underwriting requirements can preserve incentives. These are ordinary insurance-economic mechanisms, whose strength depends on contract terms and decision makers' information.
The effect of controls cannot be inferred from a simple comparison of claims among insured firms that do and do not buy coverage. High-risk organizations may be more likely to purchase the broadest insurance and to invest in stronger controls after losses. Claims are filtered by policy definitions and deductibles. These selection and measurement effects create a severe risk of mistaking an observed correlation for a prevention effect.
For insurers, external fraud can involve correlated campaigns targeting firms that use similar vendors or payment software. The discussion by Böhme and Kataria (2006) of correlation in cyber-insurance risks distinguishes within-firm dependence from dependence across insured enterprises. While their working paper studied broader information-security risks, the distinction helps organize a research question about shared fraud infrastructure. It does not provide calibrated correlations for BEC loss portfolios, and none are claimed here.
Contractual consequences of employee misconduct
A claim may involve both negligent employees and a dishonestly participating employee. The latter changes the predicate facts for an employee-dishonesty agreement; the former may be relevant to policy conditions or to the causal inquiry but cannot automatically substitute for dishonest intent. If a perpetrator later becomes an employee, the insured must establish the relevant status at the time of the act as required by the wording. If one insured entity's employee steals another insured entity's assets, coverage may depend on intercompany definitions and the treatment of property of others.
The practical research result is a requirement for actor-specific documentation: employment and contractor records, bank mandates, dual-approval controls, individual benefit flows, custody arrangements, interviews and contemporaneous communications. The claim narrative should never collapse the entire organization's agency into the phrase 'the insured sent the money.' Individual acts and contractual definitions determine whether the relevant person was a dishonest actor, a deceived decision-maker, or an uninvolved controller.
Digital Payment Loss Verification
What the payment record proves
Digital payment systems record transactional facts, but those facts answer different questions. A bank debit proves an amount was posted to an account. A payment message shows the details of an order, including the account to which the sending bank directed it. A bank confirmation can establish that the receiving institution accepted a transfer. Neither record alone proves whether the named beneficiary actually controlled the receiving account, whether the payment discharged a commercial debt, or how much of the money remained recoverable when the claim was adjusted.
The verifier should obtain, according to applicable law and availability, the original vendor invoice, purchase order, invoice revision history, approved beneficiary master-data record, original instruction, full email headers, identity logs, payment authorization history, bank payment reference, SWIFT or domestic payment identifier, receiving-bank acknowledgement, recall requests, frozen balances, restitution receipts and legal assignment of rights. Each artifact should have a provenance record and timestamps with identified time zones. Preservation protects the reconstruction from later alteration of accounts or messaging records.
A rational chronology separates five time points: the deceptive message, internal authorization, bank acceptance, ultimate credit or settlement, and confirmed recovery. Detection time may occur much later. A transfer can be booked at one moment but recalled before the beneficiary obtains usable proceeds. Conversely, a recovery request may be submitted promptly while funds have already moved through multiple accounts. Claim adjustment requires documented outcomes rather than the assumption that a recall request means recovery.
U.S. bank-customer authorization
Under UCC Article 4A-202, a receiving bank may treat a payment order as effective under specified circumstances involving an agreed, commercially reasonable security procedure, good faith and compliance with the procedure. The statute distinguishes a payment actually authorized by the sender from an order made effective by the agreed verification mechanism even if it was not authorized in the ordinary sense. Application depends on state enactment, governing agreements and payment facts. This framework directly matters to the residual loss carried by a business customer after an unauthorized order.
An enterprise's claim against a bank and its claim against a crime insurer arise from different sources of obligation. If a bank must recredit the amount, the insured's net loss may be eliminated or reduced. If the customer bears an effective order under the applicable rules, the crime policy may become the principal financial protection. Whether the insurance contract applies is still evaluated separately. A finding that a bank acted in compliance with Article 4A-202 is neither a finding that a computer fraud exclusion applies nor a universal denial of crime insurance.
Other U.S. payment systems may operate under different rules and contracts. Commercial wire transfers should not be mechanically classified using consumer electronic-funds-transfer liability limits. The nature of the account, transfer system and parties matters. This monograph uses Article 4A as a documented example of commercial payment-order allocation and does not prescribe its application to every domestic or international transfer.
European payment services and payee verification
The revised Payment Services Directive distinguishes unauthorized payment transactions from other payment-service disputes. Article 72 addresses the burden of establishing authentication and execution in an unauthorized-transaction dispute; Article 73 provides the general refund framework for unauthorized transactions, subject to the directive's conditions; Article 74 allocates specified payer liability. Whether a corporate account holder can invoke or contractually vary aspects of this regime requires attention to the directive's scope, national implementation and the parties' legal status.
An employee deceived into authorizing a payment may have given a technically valid instruction. That outcome calls for analysis of manipulation and recipient verification, rather than an automatic assertion of an unauthorized payment under Article 73. The same distinction explains the policy interest in recipient verification for authorized credit transfers.
Regulation (EU) 2024/886 introduces verification-of-payee requirements in the European instant payments framework, including comparison of the payee name or other permitted business identifiers against an account identifier. The ECB implementation schedule identifies October 9, 2025 for euro-area PSPs to implement verification of payee and July 9, 2027 for PSPs in non-euro-area member states, with institutional category-specific timelines. A verification result is a warning or matching service; it is not a guarantee that the payee has a legitimate commercial purpose or that an authorized transfer cannot be fraudulent.
The distinction between payer manipulation and unauthorized initiation is supported by the 2025 EBA–ECB fraud report. It reports EUR 4.2 billion of fraud across monitored EEA payment instruments in 2024 and states that manipulation of payers accounted for more than half the value of fraudulent credit transfers. It further reports that payment-service users bore approximately 85% of credit-transfer fraud losses in the year examined. These figures concern EEA payment fraud and liability allocation within the measured scope, not the dollar value of insured corporate crime claims.
United Kingdom: reimbursement of authorized push payment fraud
The United Kingdom demonstrates a distinct approach to the deceived but authorizing payer. The Payment Systems Regulator's APP fraud reimbursement protections took effect for qualifying Faster Payments on October 7, 2024 and apply to individuals, microenterprises and charities within the scheme's terms. The PS24/7 reimbursement decision sets a maximum reimbursement level of GBP 85,000 per claim for the specified system and rules. Ordinary larger corporate payers should not be presumed eligible merely because the transfer was fraudulent.
The PSR reimbursement dashboard reports that, over the 18 months from October 7, 2024 to March 31, 2026, 88% of money lost in in-scope APP scam claims was reimbursed, representing GBP 316 million; approximately 438,300 claims were reported and 301,500 were in the scheme's scope. These results measure the defined reimbursement regime and reporting periods, rather than loss experience across all banks or all commercial insurance policies. The distinction between claims reported and claims in scope is central.
An insurer examining a qualifying small enterprise's loss must determine what the payment provider has already reimbursed or remains obligated to reimburse under the scheme and the contractual treatment of recoveries. The bank's payment obligation may significantly alter the net amount the insurer ultimately indemnifies. This legal relationship is not equivalent to writing an insurance endorsement: it creates a separate source of recovery.
Singapore: responsibility after phishing
Singapore's Shared Responsibility Framework announcement establishes an operational framework effective December 16, 2024, assigning defined anti-phishing duties to relevant financial institutions and telecommunications providers and setting expectations for payouts where the framework conditions are satisfied. Its scope concerns specified phishing scam circumstances and account-holder protections. It should not be described as general commercial crime insurance or a universal reimbursement entitlement for all authorized payment fraud.
The international significance is the different unit of legal responsibility. The United States example assesses agreed security procedures and the bank-customer payment order. The European framework combines refund rules for unauthorized transactions and verification mechanisms addressing manipulation. The United Kingdom's APP system protects defined customer classes from a specified type of authorized scam. Singapore allocates responsibility for defined phishing exposures across regulated intermediaries. Each changes the available paths to restitution, and each must be applied with its own effective date and scope.
A source-verifiable reconstruction of a hypothetical claim
Consider a hypothetical U.S.-dollar payment by an insured manufacturer. An impostor changes the bank details in a genuine supplier invoice, and the accounts payable department transfers USD 800,000. The receiving bank later returns USD 120,000 and the payer's bank pays an additional USD 80,000 under a documented settlement. Assume these recoveries are distinct, final and attributable to the same transfer. The remaining principal loss is USD 600,000.
Assume for illustration an affirmative social engineering agreement with a USD 250,000 per-occurrence sublimit and a USD 25,000 deductible applied to the eligible principal before the sublimit. Under that assumed order, indemnity is min(max(600,000 − 25,000, 0), 250,000) = USD 250,000. If instead the deductible reduces the amount after applying the USD 250,000 limit, the result is USD 225,000. If an applicable exclusion removes coverage, contractual indemnity may be zero. These results are not alternative interpretations of one verified contract: they demonstrate the dependence of payment on specified wording and its application.
The residual enterprise economic loss after the illustrative USD 250,000 insurance payment would be USD 350,000, ignoring claim expenses, litigation, tax, timing and other contractual rights. Under the alternative USD 225,000 result it would be USD 375,000. The insured's original USD 800,000 outflow must not be added again to the residual loss. The two bank payments must not be counted both as recoveries and as separate insured claims.
The exercise is reproducible, but it carries no statistical inference. No real loss file, insurer promise or jurisdiction-specific determination underlies the chosen sublimit and deductible. Its scientific purpose is to make the relationship between cash tracing and contract application explicit.
Evidence Validation and Uncertainty
Evidence classes and limits
The research corpus contains several distinct kinds of evidence. Appellate decisions establish facts and policy interpretation as recorded in litigation; they are selected for dispute and may not represent routine claim settlements. Insurer descriptions establish products and declared market practices; they do not provide binding coverage terms for unidentified buyers. Payment-services law identifies statutory obligations within specified jurisdictions and periods. Supervisory and professional standards provide normative control and modeling guidance. Official fraud statistics describe reported events or transactions under defined measurement systems. Peer-reviewed literature synthesizes or investigates risks subject to its design and selection limitations.
The central scholarly safeguard is to preserve these categories throughout the inference. It would be invalid to treat a legal holding as evidence of claim frequency, a bank fraud transaction rate as a crime insurance premium rate, or an insurer's marketing statement as proof of an individual's policy wording. Evidence validity depends on the question each source is capable of answering.
The FBI's 2025 complaint data concern reports received by IC3, including cases that may differ greatly in legal ownership, jurisdiction, insured status and subsequent recovery. Its BEC category includes attempts and occurrences within an agency reporting taxonomy; its monetary loss totals are not an insurer-paid-loss series. Temporal comparisons require attention to reporting behavior, categories and nominal currency rather than an assumption of stable incidence.
The EBA–ECB's 2024 payment figures have a different denominator: payment transactions across specified EEA instruments and reporting entities. A tiny fraud value as a share of the enormous value of payments may coexist with substantial absolute losses. The proportion borne by customers is a loss-allocation statistic, not a policy retention or deductible. Combining those figures with the FBI totals would require harmonization that the publicly reported aggregates do not permit.
The ACFE 2026 investigated case sample measures identified occupational fraud that investigators selected under survey criteria. Its large observed losses and diverse cases help expose mechanisms of insider conduct. They cannot establish population frequency among all employers because cases are neither a random enterprise sample nor observed with uniform detection probabilities. A full statistical treatment would require employee exposure, organizational denominators, investigation selection and claims links.
Empirical model design without invented parameters
For actuarial analysis, define annual claim count N for a documented exposure unit, and event severity S after identifiable recoveries but before insurance terms. A compound-loss representation L = sum of S over N events is a familiar actuarial identity. The insured paid loss requires application of the contract's deductible, per-event limit, aggregate, attachment points, exclusions and contribution from other recoveries. Consequently, expected gross fraud loss, expected eligible loss and expected insurer payment are generally different quantities.
A model of crime frequency requires an explicit exposure denominator: corporation-years, payment counts, transferred value, employee-years or another well-defined unit. One denominator can be suitable for one mechanism and inadequate for another. Employee theft might be modeled partly through workforce and access characteristics; payment diversion may be conditioned on number and value of payment instructions, supplier relationships, geography and control maturity. Multinational firms may centralize treasury functions, so multiple insured legal entities cannot automatically be treated as independent observations.
For severity, the empirical distribution of gross diverted funds must be reconciled to frozen amounts, recoveries, refunds and ultimate financial loss. The adjustment must preserve delays and costs. A bank's recovery received several months later can reduce ultimate principal loss while leaving temporary liquidity exposure. The decision to discount recoveries depends on the valuation question and contractual terms. Fraud loss data frequently omit claims that fall below deductibles, fail a coverage trigger, or are resolved without insurance notification.
Eling and Schnell (2016), in their review of cyber insurance insurability, identify data limitations, dynamic risk and accumulation as important barriers to reliable underwriting. These are relevant here because social engineering schemes change in response to controls and exploit networks of shared technology and payment relationships. The review is broader than corporate crime insurance; it supports the nature of the modeling challenge rather than providing BEC-specific parameter values.
A transparent portfolio illustration
Suppose a hypothetical portfolio contains 10,000 homogeneous one-year exposures and has an assumed incident probability of 0.002 per exposure-year. Assume at most one incident per exposure, independent across entities, and a fixed ultimate principal loss of USD 600,000 per incident. The expected count is 10,000 × 0.002 = 20 and the expected aggregate principal loss is USD 12 million. These assumptions are illustrative, not estimates from the FBI, ACFE or EBA datasets.
Now suppose every incident satisfies a social engineering agreement with a USD 25,000 deductible and a USD 250,000 sublimit, applied in that order. The illustrated insurer payment is USD 250,000 per event; expected aggregate insurer payment is 20 × 250,000 = USD 5 million before expenses, reinsurance and taxes. This is an expected value under deliberately restrictive assumptions, not a commercial rate indication. An underwriter would need data on variation in severity, shared attacks, policy wordings, exposure selection and the probability that a reported incident qualifies for the cover.
If incidence instead equals 0.001, expected insurer payments become USD 2.5 million under the same assumptions. If severity is large enough that the per-event sublimit continues to bind, increases in severity may leave the per-event insurer payment unchanged while sharply increasing uninsured enterprise loss. If the deductible moves above the assumed residual loss, payments fall to zero. These counterfactuals explain why policy structure and loss distributions jointly determine transferred risk.
The independence assumption deserves particular scrutiny. A major fraud campaign may compromise a payment software provider, vendor email domain or authentication process used by many policyholders. A shared perpetrator can attack several insureds with correlated timing, while different internal employee frauds may be comparatively idiosyncratic. Böhme and Kataria's work on risk correlation distinguishes within-organization and cross-organization dependence. This conceptual distinction is informative, but the paper cannot validate the particular incident probability or dependence structure of the illustrative portfolio.
An adequate actuarial study would report exposure definitions, distribution fit, censoring and truncation, recovery delays, inflation, changing controls, concentration, calendar effects and scenario selection. The U.S. ASOP 23 data quality guidance and ASOP 56 modeling guidance provide professional standards for relevant actuarial work, including data limitations, model purpose, assumptions, testing and disclosure. They are U.S. professional standards and must not be presented as universally binding regulation.
Detection, reporting and causal uncertainty
Observed fraud events pass through a reporting filter. A firm may detect a suspicious payment before settlement, block it without monetary loss, recover the money quickly, settle without insurance or report a claim. An insurer may decline claims that do not satisfy definitions. A regulator may collect transactions reported by payment providers rather than individual victims. These different thresholds create missingness that is related to control strength, loss size and the incentive to report.
A source-comparison exercise must therefore distinguish reported criminal activity, verified payment fraud, legally recognized bank liability, alleged insured occurrence, accepted insured loss and actual insurance payment. Each stage is narrower for some purposes and wider for others. No single dataset examined here spans all stages with common identifiers, so the article does not estimate the proportion of global BEC loss indemnified by crime insurers.
The apparent success of preventive controls also faces selection bias. Firms with stronger controls may detect more attempted fraud, making incident counts appear higher; firms with larger losses may improve controls and buy coverage afterward; insurers may offer different wordings to applicants with better controls. A causal effect of callbacks or beneficiary matching requires suitable comparisons or designs that isolate these mechanisms. This monograph identifies the measurement requirements without fabricating effect sizes.
International source reconciliation
The international comparison yields four separate accounting questions. In the United States, for the payment orders within its domain, Article 4A addresses bank-customer allocation of unauthorized orders through security procedure rules. In the EU, PSD2 identifies unauthorized transaction claims and Regulation 2024/886 addresses recipient matching. In the UK, the APP reimbursement system adds a specific entitlement for defined customers and transfers. In Singapore, a phishing-specific responsibility framework determines the consequences of defined failures by intermediaries. None of these regimes can be merged into a single reimbursement rate or set of universal insurance defenses.
Currency, dates and units should remain attached to each source. The FBI reports U.S.-dollar complaint losses for 2025; the EBA–ECB reports euro-denominated EEA payment fraud for 2024; the PSR reports sterling refunds for defined in-scope claims through March 2026; ACFE reports detected occupational fraud case losses in its 2026 survey. Converting currencies at a spot rate would not resolve differences in event definition or population. The scientific comparison is institutional and causal, with quantified results quoted inside their proper denominators.
Editorial and methodological verification
Each documentary claim in this study is traceable to a named source and an identifiable published edition. Judicial sources are presented with their procedural status and date. Insurer product pages are treated as descriptive and potentially subject to changing marketing conditions. Statutory materials are tied to their relevant enactment and effective periods. Public aggregate statistics are not transformed into actuarial parameters. Original illustrative calculations disclose their assumptions and arithmetic. Where evidence cannot support the stronger claim, the interpretation is correspondingly narrower.
This method aligns with the IAIS approach to documented controls and established actuarial guidance while remaining an original analytical synthesis. No undocumented claims of a global market size, universal policy cover, causal benefit of a specified control, or novel mathematical theory form part of the conclusions. The unresolved empirical question is how to obtain consistently coded, permissioned claim-level and bank-recovery records that can connect policy triggers to realized losses across legal systems.
Integrated Analysis and Research Agenda
The coverage question is an evidence question
The six-part investigation produces a central, verifiable conclusion. Corporate crime coverage begins with the actual insured interest and policy wording, while its application depends on documentary reconstruction of the event. A financial loss is a measured amount; a covered loss is an amount that satisfies a legal contract. The path from one to the other requires identifying actor, authority, fraudulent act, financial transfer, causal connection and subsequent recovery. A classification based only on a label such as cyberattack or employee error omits decisive information.
The appellate comparisons demonstrate that a court may recognize direct causation despite manual employee approvals, as in American Tooling Center (2018), or find computer use too remote, as in Apache Corp. (2016). The difference is analytically useful because it forces the investigator to inspect contract language, evidentiary facts and governing law rather than treating one court's vocabulary as the definition of an insured peril.
Market design consequences
A corporate insurance buyer can assess the interaction of crime and cyber contracts by comparing actual insuring agreements side by side, focusing on voluntary-parting exclusions, funds transfer definitions, sublimits, territorial scope, treatment of property belonging to others, conditions and other-insurance provisions. Marsh's crime insurance comparison illustrates why affirmative social engineering endorsements should be read alongside other crime agreements rather than presumed to fill every gap.
For underwriting, an application should elicit the number and value of outgoing payments, the degree of delegated approval, the use of supplier master files, the economic dependence on key vendors, cross-border banking exposure and the reliability of independently verified destination changes. A nominal promise of employee training is less probative than dated evidence of control execution. Insurers may respond through coverage definition, retention, sublimit, premium, controls assistance or exclusions, but each response is a contractual and commercial decision rather than an inevitable scientific prescription.
For claims, the greatest value of a common evidence standard is comparability. An event file can preserve original principal, affected legal entity, transfer rail, criminal method, employee's actual role, bank authorization status, exact policy text, recovery timeline, final insured loss and ultimate payment. Such records make it possible to distinguish external fraud from internal theft, record partial losses accurately, and study policy outcomes without conflating declined claims with absent fraud events.
Implications for risk science
The current sources support a layered but conventional analytical approach, not a new theory. Crime exposure is caused by dishonest behavior and vulnerabilities in organizational authorization systems. Financial realization depends on payment execution and recovery. Insurance response depends on contractual triggers and legal interpretation. Observed data depend on detection, reporting and adjudication. Each step can be studied with existing legal, economic and actuarial methods, provided the evidence does not erase their differences.
The ACFE occupational fraud methodology and FBI's complaint series should be treated as complementary descriptions of different populations. Their simultaneous inclusion is scientifically valuable precisely because it exposes why a global estimate of insured fidelity and BEC losses cannot be obtained by simple addition or multiplication. A valid estimate would demand matched populations and measured insurance outcomes.
The bank-institution dimension is likely to become more important as beneficiary verification becomes routine. The ECB's implementation schedule makes clear that phased European verification obligations alter the prevention environment at identifiable dates. Future studies can test whether changes in the payment system shift the relative frequencies of attempted diversion, successful settlement, recovered funds and claims. Such research needs pre- and post-implementation observations with comparable reporting definitions.
Research directions with testable questions
First, longitudinal case-level research should measure how frequently affirmative social engineering endorsements respond to genuine employee-authorized payments across industries and jurisdictions. Essential variables include policy version, attack channel, amount, bank recovery and coverage disposition. A blinded independent rereading of a sample of contract wordings could measure classification reliability.
Second, comparative legal work should follow the consequences of a qualifying fraudulent instruction through the bank-customer contract and the insurance contract. The main challenge is temporal sequence: legal recovery rights can arise before an insurer adjusts the loss, while realized restitution may occur afterward. The claim file must preserve the timing and legal basis of each receipt to avoid counting the same recovery twice.
Third, a payment-focused actuarial dataset should observe payment exposures, detected attempts, successful redirections, gross outflows, recovered principal and accepted insurance claims. Such a dataset would permit estimation of both event frequency and payment conditional on contract terms. It would also allow assessment of changing attacker behavior after control interventions, including more plausible impersonation or substitution of payment channels.
Fourth, research on correlated campaigns should link incidents that share supplier networks, messaging infrastructure, compromised services and criminal beneficiary accounts. It should distinguish dependence arising from a common attacker from dependence produced by the same commercial control weaknesses. The Eling–Schnell research synthesis establishes the broader need to account for dynamic and accumulating cyber losses; targeted crime-claims research must establish its own frequency and severity evidence.
Fifth, research on claim fairness should investigate how firms of different sizes can document payment instructions and recoveries. Large treasury organizations may produce extensive electronic logs, while smaller firms may retain only emails and bank statements. Evidentiary access can affect time to resolution even when the underlying financial injury is comparable. This is an institutional question requiring real claim records, not a basis for assuming differential contractual rights.
Original analytical contribution and its limits
The author's contribution is an integrated, source-tested account of the relation between four domains often studied separately: insured dishonest conduct, enterprise authorization, payment-system loss allocation and contract-bound indemnity. The original synthesis reconciles the U.S. policy-causation disputes with the distinct European, British and Singaporean payment-law approaches; separates occupational, complaint and transaction datasets; and applies an independently reproducible numerical example to the consequences of coverage sublimits, recoveries and deductibles. The contribution is comparative analysis and methodological clarification. It establishes no independent doctrine, named framework, causal estimate or new pricing formula.
The principal limitations are substantive. The research has no confidential insurance policies beyond the public descriptions and court records examined, no nonpublic claims register, no enterprise-level payment exposure dataset and no full-text exhaustive survey of every national legal system. A number of scholarly sources supply accessible abstracts and published metadata rather than unrestricted underlying datasets. Courts decide litigation-selected disputes and product pages describe particular market offers; neither source class identifies the global prevalence of coverage grants. These limitations determine the permitted scope of the conclusions and the empirical agenda.
Conclusion
Corporate crime fidelity and social engineering coverage concerns the transfer of enterprise financial loss under specified contracts in the presence of dishonest acts. Its legal and actuarial analysis requires a precise distinction between what criminals did, how an organization authorized or failed to authorize payment, what a bank executed, how much money remained lost after restitution, and which insured peril the contract actually describes.
First, employee dishonesty, externally induced payment, funds transfer fraud and computer fraud are related but contractually distinguishable event classes. The factual role of an employee does not by itself determine the criminal actor or the insurance outcome. Second, the U.S. appellate decisions show that direct-loss reasoning is sensitive to policy terms, the factual chain and the governing law. Manual approval may be compatible with covered direct loss under one form, while incidental use of email may be insufficient under another.
Third, financial institutions and public payment systems influence the final insurance exposure. UCC Article 4A, European payment law and payee verification, British authorized-push-payment reimbursement, and Singapore's phishing responsibility regime create distinct rights and obligations. Their scope and effective dates must be established before treating a bank payment as a recovery against the insured's principal loss. Fourth, real-time financial tracing and independent proof of the recipient, instruction and reimbursement materially improve the quality of claims adjustment.
Fifth, the best available public measurements establish substantial reported fraud losses, but the FBI, ACFE, EBA–ECB and PSR data describe different populations and denominators. None directly supplies the annual insured claim frequency or expected covered severity for a corporation or underwriting portfolio. Actuarial work requires exposure-aligned observations, explicit selection corrections, recovery information and policy-term modeling. The hypothetical computations demonstrate contractual sensitivity and nothing more.
The general result is that sound crime insurance analysis joins contractual interpretation to payment evidence and economically coherent loss accounting. This provides a rigorous foundation for market comparison, underwriting inquiry, claims verification and future empirical research. It also preserves the limits of present knowledge: proof of a dishonest transfer, proof of bank liability and proof of insurance coverage are each independent tasks within a connected financial event.
References
Association of Certified Fraud Examiners (2026). Occupational Fraud 2026: A Report to the Nations. Survey methodology and results.
Actuarial Standards Board (2016). Actuarial Standard of Practice No. 23: Data Quality.
Actuarial Standards Board (2019). Actuarial Standard of Practice No. 56: Modeling.
Böhme, R., and Kataria, G. (2006). Models and Measures for Correlation in Cyber-Insurance. Workshop on the Economics of Information Security.
Chubb (2026). Social Engineering Fraud Insurance: Coverage for Crime Insurance.
Cross, C., and Gillett, R. (2020). Exploiting Trust for Financial Gain: An Overview of Business Email Compromise (BEC) Fraud. Journal of Financial Crime 27(3), 871–884.
Eling, M., and Schnell, W. (2016). What Do We Know About Cyber Risk and Cyber Insurance? Journal of Risk Finance 17(5), 474–491.
European Banking Authority and European Central Bank (2025). Joint Report on Payment Fraud: Press Release and Data Summary, December 15, 2025.
European Central Bank (2024–2025). Instant Payments Regulation: Requirements and Implementation Schedule.
European Parliament and Council (2015, consolidated 2025). Directive (EU) 2015/2366 on Payment Services in the Internal Market (PSD2).
European Parliament and Council (2024). Regulation (EU) 2024/886 on Instant Credit Transfers in Euro.
Federal Bureau of Investigation, Internet Crime Complaint Center (2026). Internet Crime Report 2025. FBI IC3 Annual Report, including complaint tables and recovery operations.
Info-communications Media Development Authority and Monetary Authority of Singapore (2024). Implementation of Shared Responsibility Framework, Joint Announcement, October 24, 2024.
International Association of Insurance Supervisors (2024). Insurance Core Principles and ComFrame, including ICP 8 on Risk Management and Internal Controls.
Marsh (2026). The Basics of Commercial Crime Insurance.
National Institute of Standards and Technology (2024). The NIST Cybersecurity Framework (CSF) 2.0. NIST CSWP 29.
Payment Systems Regulator (2024). PS24/7: Faster Payments APP Scams Reimbursement Requirement, Maximum Level of Reimbursement.
Payment Systems Regulator (2024–2026). APP Fraud Reimbursement Protections: Eligibility and Scope.
Payment Systems Regulator (2026). APP Scams Reimbursement Dashboard: Results through March 31, 2026.
Travelers (2026). Fidelity & Crime: Applications, Social Engineering Supplement, and Sample Policy Forms.
Uniform Law Commission / Legal Information Institute (1989, current text). Uniform Commercial Code Article 4A, Section 4A-202: Authorized and Verified Payment Orders.
United States Court of Appeals for the Fifth Circuit (2016). Apache Corp. v. Great American Insurance Co., No. 15-20499, October 18, 2016 (unpublished).
United States Court of Appeals for the Sixth Circuit (2018). American Tooling Center, Inc. v. Travelers Casualty & Surety Co. of America, No. 17-2014, July 13, 2018.
United States Court of Appeals for the Second Circuit (2018). Medidata Solutions Inc. v. Federal Insurance Co., No. 17-2492-cv, July 6, 2018 (summary order).
United States Court of Appeals for the Eleventh Circuit (2019). Principle Solutions Group, LLC v. Ironshore Indemnity, Inc., No. 17-11703, December 9, 2019.
United States Court of Appeals for the Ninth Circuit (2022). Ernst & Haas Management Co. v. Hiscox, Inc., No. 20-56212, January 26, 2022.