dissenting.
In affirming the district court’s coverage conclusion, today’s majority adopts a simplistic interpretation of the policy terms that disregards important coverage limitations. As I see it, the policy required a direct loss to First Defiance, and Hunt’s theft from non-custodial customer investment accounts does not qualify as such a loss. Furthermore, neither the policy language nor the history of fidelity coverage supports the majority’s view that the customer accounts constituted First Defiance’s “Covered Property.” I respectfully dissent.
I.
The majority presents this case as a simple matter of contract interpretation. To prevail, First Defiance must show three things: covered property, direct loss, and Hunt’s manifest intent. While I generally agree with this formulation, the policy language compels me to reject two points essential to the majority’s coverage conclusion: (1) that First Defiance’s fiduciary obligation for the customer accounts rendered them “Covered Property”; and (2) that the exception to the Brokerage Services Modification exclusion (“BSM exclusion”) supports coverage.
A. Covered Property
The term “Covered Property” includes property “owned and held by someone else under circumstances which make the [ijnsured responsible for the [pjroperty prior to the occurrence of the loss.” Under the majority’s interpretation of “responsible for the [pjroperty prior to the occurrence of the loss,” the mere fact that First Defiance undertook a prospective fiduciary duty — by promising good-faith investment services and accessing customers’ accounts with their permission — rendered First Defiance responsible for the accounts prior to Hunt’s theft. Though I accept the existence of a fiduciary duty, I do not understand its relevance under the policy language. The word “fiduciary” does not appear in the definition of “Covered Property” or any other relevant provision,1 so for the ma*275jority to be right, fiduciary status must qualify under the “responsible for” language.
This cannot be. First, as a purely textual matter, reading responsibility so broadly negates the “prior to the occurrence of the loss” limitation. Though not clear from the majority opinion, First Defiance’s fiduciary status arises by operation of law, either by statute or by common law duty; it does not flow from its investment agreements or the fidelity bond. If the word “responsible” encompasses such implied legal duties, then it does not matter whether the insured assumed such responsibility prior to the loss, as the policy requires. We should not read responsibility to swallow its prior-to-loss limitation. See Cleveland Elec. Illuminating Co. v. City of Cleveland, 37 Ohio St.3d 50, 524 N.E.2d 441, 444-45 (1988) (instructing courts interpreting contracts “to give effect to the words used” and not “delete words used” or “insert words not used” (citation, internal quotation marks, and emphases omitted)); Bondex Int'l Inc. v. Hartford Accident & Indem. Co., 667 F.3d 669, 677 (6th Cir.2011) (“We review policy terms in the context of the whole policy so as to read the policy terms in harmony.”).
This is especially so, considering how the prior-to-loss language came to be. The Surety & Fidelity Association of America (SFAA) developed the current standard definition for “Covered Property” directly in response to two federal courts of appeals decisions on the prior versions of the provision. See Vons Cos. v. Fed. Ins. Co., 212 F.3d 489, 491-93 (9th Cir. 2000); Lynch Props., Inc. v. Potomac Ins. Co. of III., 140 F.3d 622, 628-30 (5th Cir.1998). In those cases, coverage extended to property “for which the Insured is legally liable.” Like the instant case, both Vons and Lynch involved settlements between insured employers and third parties harmed by the dishonest actions of covered employees; in each instance, the employer sought fidelity coverage for the amounts reimbursed to customers due to its legal liability for the customer losses. Both courts declined to read “legally liable” so expansively, reasoning that the insured’s interpretation transforms the standard fidelity-bond policy — a limited form of coverage — into a general liability policy. Vons, 212 F.3d at 492-93; Lynch, 140 F.3d at 629.
The industry responded to these favorable interpretations in 2004 by incorporating a key distinction from Lynch into the definition of “Covered Property”: for fidelity coverage to apply, the employer’s responsibility for the loss must vest “prior to the occurrence of the loss.” See Lynch, 140 F.3d at 629 (emphasizing that “[the insured employer] does not argue ... that it was legally liable for the funds prior to their theft” in distinguishing fidelity coverage from general liability coverage); Robert J. Duke, A Brief History of the *276Financial Institution Bond, in Financial Institution Bonds 1, 28 (Duncan L. Clore ed., 3d ed.2008) [hereinafter “Duke, A Brief History ”] (explaining that the covered-property revision “codif[ied]” the Lynch interpretation). This change clarified that a fidelity bond, unlike a general liability policy, provides no coverage for an employer’s vicarious liability for employee torts. See Vons, 212 F.3d at 491-93; Lynch, 140 F.3d at 628-30; Duke, A Brief History at 27-28.2 This drafting history persuades me that the majority overemphasizes the word “responsible” at the expense of the limiting language “pri- or to the occurrence of loss.”
Without addressing the drafting history, the court argues that the whole point of this “Covered Property” definition “is to reach property not owned and not held by the insured.” No quarrel there; the definition obviously extends coverage to some noncustodial property. But the majority’s focus on function overlooks the limits of the new policy language. Compared to the “legally liable” language of its predecessor, the new definition adopted in 2004 covered no new ground by reaching non-custodial property. Rather, the new definition added a temporal element — the prior to loss language — to cement the principle recognized in Lynch: the insured’s responsibility for the stolen property must arise prior to the loss, not by virtue of vicarious liability. Nothing in the drafting history of this provision suggests that the substitution of the similar term “responsible” for the phrase “legally liable” overrides the limiting effect of “prior to the occurrence of loss.” And, accepting the existence of a fiduciary duty, neither First Defiance nor the majority presents any authority for the proposition that a prospective fiduciary duty, by itself, qualifies non-custodial customer property for coverage under this term.
Lynch, in fact, held the opposite. The fraud in Lynch resembled the theft in this case in many respects. Lynch Properties, the insured employer, orally contracted to provide investment management services to the company president’s mother. The employee/bookkeeper entrusted with monitoring the mother’s off-site financial accounts then abused her check-drafting authority to raid the accounts. Lynch Properties reimbursed the stolen funds and then sought coverage for the losses. Lynch, 140 F.3d at 624-25. Notwithstanding the dishonest employee’s breach of contractual and fiduciary duties, the court denied coverage, reasoning that
[the insured employer] does not argue ... that it was legally liable for the funds prior to their theft.... While [the employer] ... argues how it may be vicariously liable for [the employee’s theft], this argument fails to show how it was “legally liable” for the stolen property itself, that is, for the funds in [the customer’s] account. Acceptance of [the employer’s] argument would mean that *277[the] policy would cover any loss where an employee takes a customer’s property in the course of their employment responsibilities, regardless of whether the employer had any interest in the property itself.
Lynch, 140 F.3d at 629; see also Vons, 212 F.3d at 491-93 (denying coverage for investors’ losses, where investors relied on fictitious grocery-diverting transactions fraudulently confirmed by an alleged employee of the covered employer). Like Lynch Properties, First Defiance lacked responsibility under the fidelity bond for the property prior to the theft.
The court attempts to distinguish Lynch, arguing that the insured employer “had no relationship to [the mother’s] personal accounts, which she kept at other financial institutions and which had no connection to the company.” True enough, but that is not a relevant distinction. Although the terms of the investment services provided in Lynch differed somewhat, Lynch Properties and its dishonest bookkeeper had the same essential contractual and fiduciary relationship vis-a-vis the non-custodial customer accounts as did First Defiance. For a fee, Lynch Properties agreed to do the following for Mrs. Lynch: “manage ... property and investments,” “writ[e] checks to pay bills, reconcile bank account statements, and prepare] financial statements.” Lynch, 140 F.3d at 625. Consequently, Lynch Properties bore at least as much responsibility for the non-custodial customer account — including an unchecked power of the purse— as Hunt did in this case.3 The fiduciary relationship predated the theft in both circumstances, but that did not make Lynch Properties responsible for the stolen money “prior to the occurrence of loss” under the bond. If, as the court stresses, fiduciary status settles the coverage dispute, then Lynch should have come out differently. Tellingly, First Defiance makes no attempt to square its arguments with Lynch.
First Defiance could have assumed responsibility for the risk of theft prior to the loss — consistent with Lynch and the new “Covered Property” definition — by placing that guarantee in its investment agreements with customers. It did not. The court excuses First Defiance’s silence on the subject of responsibility for employee thefts, finding an implicit do-no-harm guarantee in the “understandable reality” of business. In doing so, it conjures images of disgruntled grocery clerks, pilots, and waiters poisoning, marooning, and double-charging customers. Of course, I accept the existence of such implied legal duties. And given the well-established principle of respondeat superior, I agree that those hypothetical employers would probably suffer liability for their employees’ malfeasance. But the employers’ ultimate liability for those wrongs has no bearing on the scope of the insurance coverage they purchased. For non-custodial customer property, the policy required responsibility “prior to the occurrence of loss” — that is, a responsibility stemming from something more than a subordinate’s unanticipated act. No such responsibility existed here.
One additional point requires a brief response. The majority alternatively suggests that the customer accounts qualify for coverage because First Defiance “held” the property. Not even First Defiance *278argues this, as the parties agree that third party National Financial maintained custody of the customer investment accounts. Nothing in the record indicates that First Defiance held the property in those accounts at the time of the theft, and we should not assume so.
B. Exception to the BSM Exclusion
Similarly, the court’s reliance on an exception to the BSM exclusion misses the mark. Ohio law prohibits reading a coverage exclusion (or an exception thereto) to create coverage not otherwise provided by a policy’s coverage provisions. See, e.g., Blake v. Thornton, 182 Ohio App.3d 716, 914 N.E.2d 1102, 1107-08 (2009); Fid. & Cas. Co. of N.Y. v. Seven Provinces Ins. Co., 345 F.2d 227, 229 (6th Cir.1965); Gen. Mills Inc. v. Liberty Ins. Underwriters Inc., 498 F.Supp.2d 1088, 1094 (S.D.Ohio 2007). This rule recognizes the limited purpose of coverage exclusions — that of exempting particular items from coverage. Courts examine coverage provisions to answer coverage questions, not exclusions and their exceptions. The plain text of the BSM exclusion reinforces this general rule; it expressly defers to the coverage provided in the Insuring Agreement. (R. 39, Ex. 1, Fidelity Bond at 42 (limiting the BSM exclusion and its exception with the caveat “provided such unlawful withdrawal and conversion is covered under Insuring Agreement (A)”).)
Contrary to the majority’s suggestion, this reading of the exception does not obviate it. It reaches thefts from customer accounts that satisfy the covered-property and direct-loss provisions, such as accounts held by the insured or for which the insured assumes responsibility prior to the loss. First Defiance’s non-eustodial customer accounts do not meet those standards.
II.
The majority’s broad interpretation of “Covered Property” also runs counter to key coverage limitations developed by the insurance and investment industries in the 1970s and 1980s: the direct-loss requirement and the manifest-intent requirement. The industry designed these terms to curtail expansive judicial interpretations treating fidelity coverage as a form of general liability insurance. Accordingly, the direct-loss provision limits coverage to “IHoss resulting directly from dishonest or fraudulent acts committed by an Employee acting alone or in collusion with others,” and the manifest-intent provision requires that the dishonest employee act “with the manifest intent: (1) to cause the Insured to sustain such loss, and (2) to obtain an improper financial benefit for the Employer or another person or entity.” In addition to these limitations, the industry adopted exclusions (t) and (v) precluding coverage for “damages of any type for which the Insured is legally liable” and “indirect or consequential loss of any nature,” respectively. Cf. Karen Wildau, Evolving Law of Third-Party Claims Under Fidelity Bonds: When Is Third-Party Recovery Allowed?, 25 Tort & Ins. L.J. 92, 117-18 (1989) (contemplating the practical impact of the 1976-1986 limiting revisions to standard fidelity coverage). All of these limitations appear in Progressive’s policy. (Fidelity Bond at 13, 41.)
Courts responding to the revisions adopted during this period observed a clear distinction between limited fidelity coverage and general-liability insurance. See, e.g., Universal Mortg. Corp. v. Wurttembergische Versicherung AG, 651 F.3d 759, 761-62 & n. 1 (7th Cir.2011); Vons, 212 F.3d at 492-93; Lynch, 140 F.3d at 628-30; RBC Mortg. Co. v. Nat’l Union Fire Ins. Co. of Pittsburgh, 349 Ill.App.3d 706, 285 Ill.Dec. 908, 812 N.E.2d 728, 733 *279(2004); Tri City Nat’l Bank v. Fed. Ins. Co., 268 Wis.2d 785, 674 N.W.2d 617, 623-24 (App.2003); Aetna Cas. & Sur. Co. v. Kidder, Peabody & Co., 246 A.D.2d 202, 676 N.Y.S.2d 559, 564 (1998). For these courts, “ ‘direct’ means ‘direct’ and ... in the absence of a third party claims clause, [the] policy d[oes] not provide indemnity for vicarious liability for tortious acts of [the insured employer’s] employee.” Vons, 212 F.3d at 492-93; see also Universal Mortg., 651 F.3d at 762 (“[W]hen an insured incurs liability to a third party— whether in contract or tort — as a result of employee misconduct, financial loss resulting from that liability is not ‘directly’ caused by the employee misconduct and therefore is not covered by fidelity bonds containing direct-loss language.”); William T. Bogaert & Kerry Evensen, Loss and Causation Under the Financial Institution Bond, in Financial Institution Bonds 580, 596-97 (“The phrase ‘resulting directly from’ is unambiguous and its meaning indicates a stricter standard of causation than mere ‘proximate cause.’” (internal footnotes omitted)). So construed, the bond revisions of the 1970s and 80s served to remind that, despite prior interpretations to the contrary, fidelity coverage concerns only the fidelity of the employee to the employer. See Aetna, 676 N.Y.S.2d at 565 (tracing the history of fidelity coverage to the turn of the twentieth century and recognizing that, as of 1998, “[n]othing in the history of these particular bonds, which comports with an historical understanding of what fidelity coverage is, indicates that the employee infidelity being covered as a risk could reach the employee’s dishonesty toward third parties, absent an intent to cause a loss to the employer.”).
Today’s majority disregards this historical context and glosses over one of the district court’s essential holdings: that the direct-loss requirement encompasses proximate causation. In the court’s view, “[i]f property qualifies as ‘covered property,’ and a dishonest employee steals it, the employee ‘directly’ causes the loss. It is as simple as that, and that is true under any definition of ‘directly.’ ” As stated above, I part company with the majority’s interpretation of “Covered Property.” But I also disagree with the relative priority given to that term — a policy “Condition[ ] and Limitation[ ]” — compared to the direct-loss limitation in the coverage clause. “The problem with th[at] argument is that it is founded on [a coverage definition], not the insuring clauses.” Vons, 212 F.3d at 491; cf. Nationwide Mut. Ins. Co. v. Wright, 70 Ohio App.3d 431, 591 N.E.2d 362, 365 (1990) (placing greater weight on language appearing in the coverage clause than on external limitations). The plain language and drafting history of the direct-loss requirement persuade me that the district court erred in grafting a proximate-cause standard onto the direct-loss language. Ohio’s courts must settle that issue.
III.
I do not doubt that First Defiance could be held liable for Hunt’s thefts, but this responsibility arises after the fact from its vicarious liability for its employee’s misconduct. That is not a direct loss. While general liability policies ensure against such risks, a fidelity bond does not. See, e.g., Universal Mortg., 651 F.3d at 761-62 & n. 1; Vons, 212 F.3d at 492-93; Lynch, 140 F.3d at 629; RBC Mortg., 285 Ill.Dec. 908, 812 N.E.2d at 733. First Defiance did not purchase a general liability policy. I respectfully dissent.
. Nor, for that matter, does it appear in First Defiance’s appellate briefs, other than a pass*275ing reference in its discussion of a completely unrelated policy provision. (See First Defiance Br. 46 n. 182.)
During oral argument, First Defiance suggested that its investment agreements expressly assumed responsibility for the property in the customer accounts. Given the opportunity to elaborate, it could not identify such a provision, and the record reveals no support for this position. Instead, the undisputed record reflects that, at the time they opened their brokerage accounts, the customers signed acknowledgments recognizing that the accounts differed from bank accounts in that they "are not FDIC insured” and "are not deposits or other obligations of the bank and are not guaranteed by the bank.” First Defiance reminded its customers of these risks in its periodic account reports. While these disclaimers did not tell customers that they would assume the risk of employee theft, they clearly did not tell customers that First Defiance assumed that responsibility, either.
. At the same time that the industry updated "Covered Property,” it revised other bond terms to distinguish fidelity coverage from general liability insurance. Duke, A Bnef History at 21 (explaining that the 2004 revisions sought to "correct continuing misinterpretations of the bond by some courts”), 21-22 (noting the replacement of the manifest-intent requirement — construed by courts to encompass non-purposeful conduct — with an "active and conscious purpose” standard requiring purposeful conduct), 27 (addressing a modification to exclusion (t) that, along with the other revisions, "remove[d] any suggestion that the legal liability of the Insured has any effect on the obligations of the Underwriter”); see also Stephen M. Kranz et al., Can Direct Mean Direct? Untangling the Web of Causation in Fidelity Coverage, 16 Fid. L.J. 153, 154-56 (2010) (characterizing the new definition of "Covered Property” as a further industry attempt to narrow fidelity bond coverage, akin to the adoption of the direct-loss requirement).
. The investment agreements in this case required a third party to serve as custodian of the accounts and reserved ultimate investment discretion for the customer. (R. 46, Ex. D, Investment Manágement Agreement § 1, 4.) As I read it, Hunt’s discretion extended no further than recommending investment decisions keyed to the client’s financial circumstances and executing investment transactions within customer-approved parameters.