dissenting.
The court’s opinion today unnecessarily raises the cost of cooperating with a government investigation. For the court, the existence of a government investigation exception to the third-party waiver rule is an impediment to the truth-seeking process. Op. at 304. After all, under the court’s rule more participants in the criminal and civil justice systems have access to privileged information, and the courts’ task of making accurate factual determinations is eased. Realistically speaking, the choice before this court today is not between narrower and wider disclosure, but between a disclosure only to government officials and no disclosure at all. Because I am convinced that a government investigation exception to the third-party waiver rule would increase the information available over that produced by the court’s rule and would aid the truth-seeking process, I respectfully dissent.
The court undertakes an exhaustive review of the other federal decisions which have addressed this question. The court’s analysis of the extant law is largely accurate. I would make clear, however, that the authority arrayed in favor of the court’s rule is not overwhelming. It is true that only one circuit court of appeals has implemented a government investigation exception to the third party waiver rule. See Diversified Indus., Inc. v. Meredith, 572 F.2d 596, 611 (8th Cir.1978) {en banc). Yet it is equally true that one other circuit court of appeals has expressly contemplated a government investigation exception where, as here, the holder of the privilege information executes a confidentiality agreement with the government before disclosure. See In re Steinhardt Partners, 9 F.3d 230, 236 (2d Cir.1993). One circuit court of appeals has rejected a government investigation exception in the case before it, but was not presented with and did not specifically comment upon cases in which the disclosing party had entered into a confidentiality agreement. Permian Corp. v. United States, 665 F.2d 1214 (D.C.Cir.1981). Only one court of appeals has rejected the government investigation exception when the disclosing party had entered into some kind of confidentiality agreement with the government. Westinghouse Elec. Corp. v. Republic of the Philippines, 951 F.2d 1414, 1427-27 *308(3d Cir.1991).1 All of the other circuit decisions that the court cites either concern whether disclosures to one federal government agency waive privilege as to another federal government agency, United States v. Massachusetts Inst. of Tech., 129 F.3d 681, 686 (1st Cir.1997) (holding that disclosure to Department of Defense audit committee without a confidentiality agreement waived privilege in IRS investigation), or address privileges other than the attorney-client privilege. See Dellwood Farms, Inc. v. Cargill, Inc., 128 F.3d 1122 (7th Cir.1997) (considering waiver rules concerning the government investigatory privilege). Needless to say, the circuit courts of appeal are deeply split on whether a disclosure of privileged information to the government, in the course of an investigation and with a confidentiality agreement, waives the privilege as to all other parties. More certainly, this court has never addressed this question, and all of the authority cited, to the extent it provides any answer, does not bind our resolution of this case.
I would have resolved this open question by holding that there is a government investigation exception to the third-party waiver rule. I address the existence of this exception with regard to the attorney-client and the attorney-work-product privileges separately below.
A. The Attorney-Client Privilege
I am unpersuaded by the court’s reasoning for its rejection of the exception. First and most generally, the court claims that the “attorney-client privilege [is] narrowly construed” and suggests that we should start with a presumption against Columbia’s claim of privilege. Op. at 293-94. It is certainly true that the application of the privilege to certain communications is to be “narrowly construed,” in part because it “reduces the amount of information discoverable in a lawsuit.” United States v. Collis, 128 F.3d 313, 320 (6th Cir.1997). Neither the parties nor the court denies that the privilege would ordinarily cover the information sought in this case. Once we decide that the attorney-client privilege applies to certain communications, the question becomes one of waiver. When the question is whether the attorney-client privilege is waived by certain actions, the presumption shifts in favor of preserving the privilege. As this court has recently made clear, “a court should begin its analysis with a presumption in favor of preserving the privilege.” In re Perrigo Co., 128 F.3d 430, 440 (6th Cir.1997).
Second, the court suggests that the exception is “unrelated to” the justification for the attorney-client privilege, that is, encouraging “frank communication” between attorney and client. Op. at 302. Therefore, the court contends, the exception ought to be rejected.
It is not clear why an exception to the third-party waiver rule need be moored to the justifications of the attorney-client privilege. More precisely, we ought to seek guidance from the justifications for the waiver rule to which the exception is made. Those justifications are not exactly coincident with the justifications for the privilege itself. Although the philosophical pedigree of the rule is unclear, I can discern at least two frequently articulated justifications for the third-party waiver rule.
The first justification draws on the reasons behind the attorney-client privilege itself. The attorney-client privilege is de*309signed to foster frank communication between attorney and client. See Commodity Futures Trading Comm’n v. Weintraub, 471 U.S. 343, 348, 105 S.Ct. 1986, 85 L.Ed.2d 372 (1985). Although some view privileges as impediments to the truth-seeking process, the calculation is that the attorney-client privilege improves the adversarial process without a net loss in the amount of information produced. Insofar as the existence of the privilege creates the communication sought, the exclusion of privileged information conceals no probative evidence that would otherwise exist without the privilege. The absence of the communication would leave the adversarial process with no more information and with counsel less able to present focused arguments to the courts.
Courts explaining the waiver rule note that once a client or his lawyer voluntarily reveals privileged information, the basic justification for the privilege no longer obtains. The disclosure indicates that the privilege was not a necessary incentive for the privileged communication to occur, by demonstrating that “the client does not appear to have been desirous of secrecy.” Permian, 665 F.2d at 1220. See also 8 Wigmore on Evidence § 2311, at 599 (1961). Courts employing this explanation for the waiver rule essentially make a statistical inference: “if clients themselves divulge such information to third parties, chances are that they would also have divulged it to their attorneys, even without the protection of the privilege.” Westinghouse, 951 F.2d at 1424.
The court’s opinion also hints strongly at this type of reasoning, suggesting that the “uninhibited approach” of recognizing an exception “has little, if any, relation to fostering frank communication between a client and his or her attorney.” Op. at 302. The analysis of these courts does not account for the element of time. Clients do not communicate with their attorneys with perfect knowledge of the future. Without the premise of perfect, or at least very good, predictive information in the hands of attorneys and their clients, there is no basis for the inference from a later disclosure to the motivation behind the privileged communication. That a client is willing to disclose privileged information to the government at time T2 indicates very little indeed about whether she would have communicated with her attorney, absent the promise of the privilege, at time Tl. In the meantime, the client certainly has learned more about intervening events and perhaps has become more legally sophisticated (through the informed legal advice arising from her candid communication with her attorney). It seems clear to me that ex ante the attorney-client privilege is generally quite important in producing the communication, and that later disclosures provide only the weakest evidence to contradict that conclusion.2
A more pragmatic approach would evaluate the impact of the waiver rule on the client’s incentives at the time of the putative disclosure. I introduce now, and will detail later, what I believe to be an uncontroversial behavioral prediction: Faced with a waiver of the attorney-client privi*310lege over the entire subject matter of a disclosure and as to all persons, the holder of privileged information would be more reluctant to disclose privileged information voluntarily to the government than if there were no waiver associated with the disclosure. This prediction raises at least one question: Is the effect on governmental investigations the type of concern of which the waiver rule may take account?
The court suggests that the proper analysis of the waiver rule cannot include mere “public policy” considerations like the efficient dissemination of information to the government. Op. at 302. Yet, the court’s analysis ignores the common law nature of the privilege inquiry in the hands of courts. The construction and interpretation of privileges, including the circumstances under which they are waived, are conferred to the “reason and experience” of federal courts. Fed.R.Evid. 501. In exercising this authority, federal courts have regularly analyzed whether particular rules are “in the public interest,” or whether the rules regarding the privilege would have undesirable side effects. These questions of “policy,” like the deleterious impact of a waiver rule on government investigations, are at the heart of the privilege inquiry. See Jaffee v. Redmond, 518 U.S. 1, 116 S.Ct. 1928, 135 L.Ed.2d 337 (1996) (engaging in a extensive “policy” inquiry to formulate a psychotherapist-patient privilege under federal law). I can find no rule narrowly constraining the considerations that courts may take into account in developing rules regarding a common law privilege or requiring that courts turn a blind eye to the practical effect of the privilege rules that they are charged to create.
Moreover, the second justification generally offered for the third-party waiver rule — preventing the selective invocation of the privilege — contradicts the court’s premise. Reducing the client’s choices to two, the complete abandonment of the privilege or preserving total confidentiality, the third-party waiver rule prevents parties from strategically deploying the most favorable privileged material while jealously guarding the most damaging. See United States v. Workman, 138 F.3d 1261, 1263-64 (8th Cir.1998); United States v. Rakes, 136 F.3d 1, 5 (1st Cir.1998). To the extent that a party seeks to use otherwise privileged evidence, his opponent should be able to respond with the same. Without the waiver rule, the worst of all circumstances could emerge: the court may be presented with an incomplete view of the facts, where exposed evidence would be contradicted by concealed privileged information.
Preventing the distortion of the record only justifies the topical scope of the waiver. By waiving the privilege as to the entire subject matter of the disclosure, United States v. Collis, 128 F.3d 313, 320 (6th Cir.1997), the third-party-waiver rule has the effect of preventing an incomplete presentation of privileged information in a particular proceeding.
Of particular concern here is the feature of the waiver rule that eliminates the privilege not only with regard to the adversary to whom the information is revealed, but also with regard to all other parties. The policy justification for this feature is much more conclusory. As the court notes, the holders of a privilege ought not have the capacity to select among their opponents. Op. at 302; In re Permian Corp., 665 F.2d at 1221 (“The client cannot be permitted to pick and choose among his opponents, waiving the privilege for some and resurrecting the claim of confidentiality as to others....”). Implicit in this feature is some conception of fairness, that the privilege is not an entitlement to discriminate between parties who are roughly on the same footing. See Op. at 303. See also In *311re Grand Jury Proceedings, 219 F.3d 175, 183, 185 (2d Cir.2000). Yet there is no reason why this amorphous appeal to “fairness” should not yield to an important public policy interest in easing governmental investigations. •
The preference against selective use of privileged material is nothing inore than a policy preference, and really also has very little to do with fostering frank communication between attorney and client. The question for this court is one of policy: Whether the benefits obtained by the absolute prohibition on strategic disclosure outweigh the benefits of the ihformation of which the government has been deprived by the rule? As the harms of selective disclosure are not altogether clear, the benefits of the increased information to the government should prevail.
The court’s rule does nothing more than increase the cost of cooperating with the government. The third-party waiver rule, if enforced in. disclosures to the government, would require an investigated party to withhold the requested information, lest she lose the privilege entirely. While it is hard to say exactly how high the marginal costs of the waiver would be without facts in an individual case, the scope of the waiver sheds some light on the general magnitude of the costs. Under the current operation of the rule, the holder of the privilege waives it as to everyone, when he discloses privileged information to a third party. Plus, the waiver covers not just the documents disclosed, but all privileged documents “pertaining to the subject matter of the disclosure.” See, e.g., Collis, 128 F.3d at 320; PaineWebber Group, Inc. v. Zinsmeyer Trusts Partnership, 187 F.3d 988, 992 (8th Cir.1999). These features combine significantly to penalize the revelation of privileged information to the government. Relatively narrow cooperation with ■ the government in the form of a disclosure of privileged information can expose an individual or firm to massive liability and reveal privileged documents far afield from the disclosure itself.
Contrary to the court’s argument, increased access to privileged information increases the absolute efficacy of government investigations, regardless of increased investigatory costs to the government. There is some evidence provided by privileged information for which there is no non-privileged substitute or to which there is no path without the privileged evidence. The court, as well as other courts addressing this question, argues that the government has “other-means” to secure the information that they need, while conceding that those other means may consume more government time and money. Op. at 303; Massachusetts Inst. of Tech., 129 F.3d at 685. Presumably, the court is referring to search warrants or civil discovery. It should be emphasized, however, that the government has no other means to secure otherwise privileged information. That the documents or other evidence sought is privileged permits the target of an investigation to refuse production through civil discovery, to quash any subpoena duces tecum, or to prevent the admission of the privileged information even by the government. The only way that the government can obtain privileged information is for the holder of the privilege voluntarily to disclose it. The court’s argument about the adequacy of other means, suggesting that the only difference between them and voluntary disclosure is cost, requires the premise that all privileged information has a non-privileged analogue that is discoverable with enough effort. That premise, however, does not hold.
Why should we minimize the cost and maximize the accuracy of government investigations? After all, as the court notes, *312private litigants also seek the truth and could benefit from the decreased costs of discovery and the increased accuracy of their positions and arguments. Op. at 303. They too, through the adversarial process, serve the truth-seeking mission with which courts are charged.
The government’s investigations are generally more important. Government officials, with finite litigative resources and no individual monetary stake in the outcome of litigation, generally are more selective regarding the matters they choose to pursue than are private parties. Because of these incentives, government investigations are more likely to be in the public interest. Private litigants, often encouraged by large potential liability, on balance will have a greater incentive to press the legal envelope and to pursue legal actions less certainly within the public interest.
The differential remedies available to public and private parties also reflect the greater importance of government investigations. The government has the authority to seek imprisonment and punitive fines. The costs and benefits of government investigations are diffuse, and therefore managing those costs and benefits most efficiently is definitionally in the public interest. On top of all this, government investigators and prosecutors start at a tactical disadvantage to private plaintiffs given the procedural protections afforded criminal defendants against the government, such as a higher burden of proof and the privilege against self-incrimination. I am comfortable, therefore, providing a clear exception for government investigations, and leaving private litigants out.
Throughout its opinion, the court suggests that recognition of an exception would deprive private parties of evidence to which they would otherwise be entitled. See Op. at 303 (characterizing the exception as a government investigatory tool “which effectively prevents future litigants from obtaining the same information”). It is important to identify the silent premise of the court’s argument: private parties would disclose privileged material to the government regardless of the existence of an exception. If the holders of the privilege did not disclose privileged information to the government, the material would still be protected by the privilege. In short, the choice presented in this case is not one whether or not to release privileged information to private parties that has already been disclosed to the government, but rather one to create incentives that permit voluntary disclosures to the government at all. In the run of cases, either the government gets the disclosure made palatable because of the exception, or neither the government nor any private party becomes privy to the privileged material.
The court finally makes a type of moral argument. Why should the government sully its hands, the court asks, by assisting in “obfuscating the 'truth-finding process’ ”? Op. at 303. The government is not about cover-ups, rather it should “act to bring to light illegal activities.” Ibid. I wonder what exactly the court thinks the government would be doing if permitted to encourage voluntary disclosure through confidentiality agreements. The government either could use the information to find additional evidence or could present the privileged information if it decided to initiate a criminal prosecution or civil action. In any event, the court’s argument misses the mark. It is not the government’s confidentiality agreement that shields the information from civil discovery by private parties, but instead the privilege itself. Without the exception, much otherwise disclosed material would stay completely in the dark, under the absolute cover of privilege. The exception aids the *313government in bringing violations of the law to light.
The theoretical merits of the exception aside, the court questions the administra-bility of the rule that it rejects. Op. at 303. In essence, the court suggests that the exception lacks rule-like features and that, as a result, the exception would dramatically increase judicial decision and private litigation costs. At first glance, it not entirely clear why the exception at issue would be more standard-like, threaten certainty, and raise decision costs. It would only apply when the disclosure was made to a government agency. It seems pretty simple to know what government agencies are. We might place a confidentiality agreement requirement on the exception.3 Really, the exception seems clear and predictable.
The court’s argument is more sophisticated, however. It claims that it would be difficult to distinguish between the government and several types of private parties who act particularly in the public interest. Op. at 303. A qui tam plaintiff, for example, brings a suit in the government’s stead. See 31 U.S.C. § 3730(b)(1). More tenuously, a plaintiff in a shareholder derivative action seeks recovery for the corporation, at least in addition to himself. The claimed risk of attempts to stretch the exception seems tenuous to me. At least the Eighth Circuit has had the exception in place for twenty-four years, see Diversified Indus., Inc. v. Meredith, 572 F.2d 596, 611 (8th Cir.1978) (en banc), and I can find no case in which a litigant has pressed the exception in these situations. Nevertheless, the quasi-governmental actors listed here seem categorically different. The difference between these actors and the government is that they stand to recover personally, and the prospect of personal recovery changes the incentives for bringing and prosecuting their actions in the first instance. See 31 U.S.C. § 3730(d) (providing substantial awards to qui tam plaintiffs if they recover for the United States). In short, because of the potentially massive recovery that both qui tam and derivative suit plaintiffs stand to receive personally, the assurance that their activities are as much in the public interest as government agency conducted investigations seems absent.
Plus, the exception is as rule-like as this court makes it, and thus the cure for this defect lies in our own hands. If the jurisprudence that this exception would produce contains a series of “difficult and fretful” exercises in “linedrawing,” Op. at 303, this court would have no one to blame but itself.
Because of a similar concern for the preservation of rule-like values in the field of privilege, I would make the exception categorical. There would undoubtedly be some voluntary disclosures to the government that would occur without the exception. The benefits of such a, disclosure to the government may, in some cases, be so great for the private party that they would outweigh the massive costs of a full subject matter waiver of privilege.4 Nevertheless, *314a rule removing the penalty of waiver from all disclosures of privileged information to the government would provide the certainty necessary to encourage cooperation with the government.
Another problem with the rule-like features of the exception is that the exception may have limited efficacy absent uniformity among courts. It would be difficult to remove the disincentive to cooperate with the government if protection from waiver depended on the circuit in which a party would be eventually involved in litigation. The mere split between our sister circuits should not dissuade us from adopting the exception. This court should follow the legal position that it finds most meritorious and leave the problem of uniformity to a higher court.
B. Waiver of the Attorney-Work-Product Privilege
The court also holds today that Columbia’s voluntary disclosure to the government of materials protected by the attorney-work-product privilege waives that privilege as well. Op. at 306-07. I believe that the reasons why the attorney-client privilege should not be waived by a disclosure to the government are also sufficient to bar waiver of the attorney-work-product privilege under similar circumstances. Some courts have decided that the requirements for waiver of the attorney-client and the work-product privileges are different. See, e.g., Permian Corp. v. United States, 665 F.2d 1214, 1219 (D.C.Cir.1981). Indeed, courts have explicitly held that “while the mere showing of voluntary disclosure to a third person will generally suffice to show waiver of the attorney-client privilege, it should not suffice in itself for waiver of the work product privilege.” United States v. AT & T, 642 F.2d 1285, 1299 (D.C.Cir.1980). Unlike the attorney-client privilege, the work-product privilege is explicitly protected by a federal rule of civil procedure, which makes no provision for waiver of the privilege. See Fed.R.Civ.P. 26(b)(3).
The court contends that the only difference between the two standards is that attorney work product may be disclosed to non-adversaries, through the so-called “common interest exception,” without waiving the privilege. Op. at 306-07. Other courts, however, have explicitly rejected the claim that the two waiver standards are identical, except for the “common interest exception.” For example, the District of Columbia Circuit has held that “the purposes of the work product privilege are more complex, and they are not inconsistent with selective disclosure' — ■ even in some circumstances to an adversary.” In re Sealed Case, 676 F.2d 793, 818 (D.C.Cir.1982).
In any event, I would also hold that the public interest in easing government investigations counsels against holding the attorney-work-product privilege waived when the holder of the privilege discloses privileged information to the government. Because I believe that Columbia intended to preserve both the attorney-client and attorney-work-product privileges and that a limited disclosure pursuant to a government agency’s investigatory request ought not waive the privileges as to all other parties, I respectfully dissent from the court’s affirming of the district court’s order compelling discovery of the otherwise privileged material in question.
. Even in Westinghouse, the court doubted that the disclosing party actually had entered into comprehensive confidentiality agreement fer all of its disclosures to government agencies. See Westinghouse, 951 F.2d at 1427.
. The court also contends that the attorney-client privilege has nothing to do with protecting communication between “a client and the government.” Op. at 302. Of • course, no one is contending that communications between the government and a private individuals are protected by the attorney-client privilege. The only situation in which the attorney-client privilege would be relevant is when a private individual discloses already privileged information to the government. The question then is whether the communication between the government and the holder of the privilege waives the already existing privilege. To me, the court’s argument seems inapposite.
. A word about the relevance of the confidentiality agreement seems appropriate. I agree with the other courts addressing this issue that parties cannot create a privilege against civil discovery by mere written agreement. See Westinghouse, 951 F.2d at 1426. The relevance of the confidentiality agreement, however, is as evidence that the holder of the privilege intended to preserve the privilege against all parties other than the government. Without such a confidentiality agreement, we could imply from the disclosure a lack of concern about the privilege.
. Or, in some cases, the costs of subject matter waiver just may not be so high that it is not a significant deterrent to cooperation with the government.