Causes of Inadvertent Disclosure of Privileged Documents and Strategies for Protecting Privileged Documents
Whitepaper
TASA ID: 1793
There is an inherent difficulty in keeping privileged documents and especially work product in e-discovery. This problem can cause the most serious damages to litigants, and consequences including anything from direct loss of cases to protracted procedural complications are reflected in a large number of cases. Each time when a privileged document is leaked, there is no real remedy. Clawing-back documents gives the producing party only a right to stop the receiving party from using the leaked documents directly. However, nothing can stop the receiving party from collecting information from other sources to strengthen its case by using the leaked documents as road maps. One should expect that the receiving party will keep copies of leaked documents for reference and may use them to cross-check anything affecting its case. Therefore, this problem must be addressed seriously. I will discuss what the main cause for the frequent leaking of privileged documents.
A Foundational Problem in the E-discovery Review Model
An inherent flaw in the network-based document review model is mainly responsible for leaking of privileged documents. This problem has existed since the start of using the network-based review model.Reviewing a pile of documents by attorneys in an offer is different from reviewing documents over the internet. The document allocation affects how documents are reviewed. This review industry has used the network-based document review model without examining its problems.
1. The network-based review model
When many reviewers review documents using a server and multiple client computers, documents are assigned to different reviewers in batches. Since different documents carry different kinds of information, the reviewers acquire different parts of the case knowledge. Therefore, they will understand the same documents in different ways, depending upon what they happen to know. Assuming that three information units A, B, and C carried in documents A, B, C are assigned to three reviewers X, Y, and Z, the three reviewers acquire different pieces of information and thus will make fatal errors, as shown below:
If Info A is essential for understanding info B in Document B, reviewer Y will make a mistake in coding for the document B. For example, if Info B indicates an email of John Doe with unknown identity, Info A is an agreement showing that John Doe is an attorney for the client. Reviewer Y will make a mistake in coding for privilege because he could not access info A. There is an unlimited number of this kind of information-information interactions in the real world. Sometimes, knowledge required for correctly coding a document may be found in multiple documents.
In business documents, nearly all names, product names, transaction names, facts and legal issues are mentioned without providing details of their legal significance. The details often appear in one or more other documents which might be non-responsive. Therefore, reviewers have to make guesses and often make wrong guesses.
2. Problems shown in an exemplar document
Corporate documents are not written with sufficient details to allow outsiders to understand. Documents are written for people who understand business, company history, people, products, and events. This can be shown in an exemplar document containing one statement: “Dear Jack, I have just signed the agreement. I will give you a copy when I get their signatures.”
How to understand this document? One cannot fully understand it even though every word is generally understood. “A word is not a crystal, transparent and unchanged, it is the skin of a living thought and may vary greatly in color and content according to the circumstances and the time in which it is used.” Justice Oliver Wendell Holmes Jr., Towne v. Eisner, 245 U.S. 418, 425 (1918).
This document cannot be accurately coded. The reviewer must make arbitrary assumptions about the agreement, the recipient, and the signers (all of those pieces of information are not provided). The same document may be a piece of junk email if the agreement is a house contract sent to a family member; the document reflects a civil violation if the agreement is an antitrust agreement sent to a partner; the document would reveal a criminal act if the agreement concerns a criminal act. A document may contain one to several terms, and even tens to hundreds of terms that are susceptible to different interpretations. Misunderstanding of many terms will not necessarily result in a coding error; however, misunderstanding one or a few critical terms may result in fatal mistakes. Due to an overwhelmingly large number of instances involving interpretation uncertainty, the total number of fatal errors is still very common. That is why the work products of human document reviewers are “full of human errors.”
3. Review model uses resources to achieve bad ends
The model flaw is responsible for massive duplicate tasks. If the review project has T facts or concepts that must be correctly understood by N reviewers, each of the reviewers have to repeat the same task to understand each of the T facts or concepts. In other words, each task is repeated by N times. The total number of tasks is N*M while it should be T tasks. In other words, the current review model wastes N-1 times efforts. If the document review lasts a long period of time, each reviewer might have to do duplicate works even for an identical task because the reviewer uses his knowledge at different times. Since the reviewer's knowledge changes by reading documents and acquiring case facts, the reviewer may make different decisions. The reviewer might repeat an identical task on the first day, the tenth day, and the fifteenth day. The reviewer might code for an identical issue differently due to changed knowledge and/or changes in memory.
This duplicate work process does not help the client in any way because it inevitably results in a massive number of conflicting coding decisions. Similar documents are often coded as both responsive and non-responsive; similar documents may be coded as both privileged and non-privileged; similar substances may be coded for redaction or for releasing. One reason for making a massive number of conflicting coding decisions is that most business documents contain large numbers of terms, persons, transactions, legal issues, background, and related events without details and documents reviewers are not intended readers who can understand every term in documents. Document reviewers must make a guess whenever they cannot access details. That is what I call “coding documents by guessing.”
4. Reviewers cannot identify many kinds of privileged documents
Frequent guesses in reviewing documents is the root cause of all sources of coding errors and inconsistencies. Naturally, it is impossible for all reviewers to accurately determine (1) work-product that does not include proper privilege stamps or contain incredible privilege stamps, (2) privileged communications without showing any attorney names, (3) improperly marked privileged documents which are concerned with innocent and “trivial” subject matters, (4) privileged documents showing confusing, undisclosed, or incredible attorney names, (5) privileged documents showing recipients that appear to be privilege-breakers, (6) improperly marked privileged documents that have been found in ordinary files, (7) improperly marked privileged communications that have been sent to business partners and agencies whose identities are unknown, unclear, or confusing to reviewers, and (8) work products such as raw test data, factual reports, and case analysis reports which have been prepared by employees acting on behalf of undisclosed counsel.
The biggest risks are reviewers' inability to recognize the client's sensitive and non-relevant information. Under the review model, the reviewers are unable to determine trade secrets, sensitive business data, customer information, and harmful information that could be used by other persons or entities to harm the client business. Such information can leak out by one or several pieces, which can be combined to gain full meaning by those who have special knowledge.
The current review model has a self-degrading feature that review accuracy degrades as the review progresses. Document context deficiency inevitably becomes worse and worse, as reviewers removed more and more non-responsive documents from the review pool. Subsequent review, second-level review, third-level review, and high-level quality review often are conducted in a review context being worse than in the first review. The tenth-round of review could be done with higher risk of errors than the first pass review because the first-pass reviewers are in a position to see more helpful information. It is not strange that a review decision is changed back and forth with no predictable improvement. This also causes a huge waste: a law firm may conduct ten rounds of review, the coding performance in the tenth round may be worse than that in the first review while the intermediate reviews keep flipping coding decisions.
Privileged Documents Leaking Reflected in Reported Cases
Bad review quality has been known since the network review model was used. I will discuss some leading cases revealing privilege disputes.
1. Privileged document leaking happens frequently
Table 1. Privileged documents produced in a few real casesNow, it is fair to say that privilege document leaking is the most common disputes in e-discovery. Similar cases can be found overall on the internet. One should note that those cases are reported only because the parties actually litigated disputes. Each time when privilege documents are produced, the producing parties have incentive to keep silence, hoping that receiving parties will not note them. This may frequently happen in large cases where a huge number of documents are produced. Most leaked documents are not known.
Those cases indicate that the impact of lost privileged documents is severe. As the court states: “loss of the attorney-client privilege in a high-stakes, hard-fought litigation is a severe sanction and can lead to serious prejudice.” Rhoads, 254 F.R.D. at 227. Some cases show that the receiving parties may immediately change their legal theories.
It is inherently a difficult task to keep privileged documents. In Ergo Licensing, Ergo reviewed documents three times: once by an associate, once by a law firm, and once more by an associate. They could not find the error until when leaked documents were used by the opposing party in a court filing. J-M Manufacturing indicates the same problem: even when the review was intended to capture leaked documents, the same error was made.
Those cases also show that using search keys are not reliable methods for keeping privileged documents. In Mt. Hawley Ins. Co., the court found the leaking party “failed to perform critical quality control sampling to determine whether their production was appropriate and neither over inclusive nor under-inclusive.…” In Rhoads Indus, Rhoads relied upon search keys to exclude privilege. This kind of mistake has been made repeatedly, but should be avoided.
The most difficult task is identifying work product. In Coburn, the leaked documents concern "underlying facts regarding the residences of certain Whitecap investors." The court in Coburn stated: “Unquestionably, reviewing documents for work product can be challenging because sometimes there are subtleties to the determination.” Nobody knows how to keep this class of documents. Law firms, in-house counsel, document processing companies can only hope that bad things will not happen. It did happen as in Cohurn and J-M Manufacturing.
The leaking party generally has only one chance to get half a remedy. In
Rhoads, upon being notified privileged documents, Rhoads re-reviewed 78,000 documents and identified 812 privileged documents on a new privilege log. The log was produced to defendants on June 30, 2008. The court states that any documents that are not placed into this review must be produced. The court essentially gave Rhoads only one chance. If it still missed privileged documents, there would be no more claw back. This rule seems to be very consistent among courts.
2. Adverse impacts of leaked privileged documents
After privileged documents are produced, clawing back privileged documents will never undo all damages. The very acts of demanding for the return of privileged documents would alert of receiving parties of the need to examine the documents to be returned. This would allow the receiving party to take notes on the content of the documents. It can affect the receiving party's litigation strategies, resources allocations, and its focus in litigation. For example, the receiving party may try to establish the same facts suggested by the privileged documents by collecting factual elements from non-privileged documents. In addition, the privileged documents may serve a road map to guile the receiving party to take certain course of actions. Privileged documents may also be used as powerful information to discourage witnesses from making conflicting testimony.
3. Producing party must take “reasonable steps” to avoid waiver
Where a party wants to get back privileged documents, Rule 26 must be considered in tandem with Federal Rule of Evidence 502(b), which provides that when a privileged document is accidentally disclosed, the disclosure will not act as a waiver of the privilege so long as the disclosure was “inadvertent” and the holder of the privilege “took reasonable steps” to prevent disclosure and to rectify the error. This rule applies only to attorney-client privilege and work product doctrine and not to other kinds of privileges. See Fed. R. Evid. 502(a).
Whether “reasonable steps” sufficient to avoid waiver have been taken is a highly fact-intensive inquiry that often leads to unpredictable outcomes. While a “reasonable steps” analysis is guided by the substantive law of the controlling jurisdiction, most jurisdictions follow the middle ground approach suggested by the committee notes to Rule 502(b). In deciding the waiver, the court considers many factors, including: (1) the “reasonableness” of the precautions taken to prevent inadvertent disclosure, (2) the number of inadvertent disclosures, (3) the extent of the disclosures, (4) any delay in measures taken to rectify the disclosure, and (5) the ever amorphous “overriding interests in justice” factor. See, e.g., Victor Stanley, Inc. v. Creative Pipe, Inc., 250 F.R.D. 251, 259 (D. Md. 2008).
The current network-based review model will effectively prevent the producing party from clawing back inadvertently produced documents for two reasons. This review model, as it stands in the current use condition, can produce privileged documents as a matter course. After the model flaw is known since 2013, a producing party cannot keep using the same flawed review model on one hand, and then argues that it has taken reasonable steps to protect privileged documents on the other hand. In other words, the common knowledge of the model problem and available solutions would effectively bar producing parties from clawing back.
However, the biggest problem is the delay in finding privileged documents. Most parties lost the right to claw back due to a long time delay in demanding for return of privileged documents. It is often difficult for document reviewers to determine privilege at the time of review. After the review is concluded, it is even more difficult for attorneys to see what is privileged. In those cited cases, it is a common scene that the attorneys for the producing parties could not know privileged documents until many months later, often when they were informed of such documents or when they saw such documents in court filings.
Such long delays are also caused by the review model. When documents are coded by a plurality of reviewers, each of who is coding documents by relying upon stories on the face and largely by guessing, they never know which privileged documents are produced and how many are produced. Search designed to cross-checking documents against privilege log is not effective. For example, search would not determine (1) documents produced by scanned images, (2) documents containing restarted privileged substances, (3) privileged materials expressed in different terminologies or out of context, (4) documents with encoding problems in foreign languages, (5) documents containing paraphrased statements which could not be caught by search keys, and (6) documents containing non-text substance such as drawing, charts, and video representation. Poorly selected search keys and invalid search algorithms also create great uncertainty. A search method, which may be excellent in one case, may be completely unworkable in other cases.
Damages to Client Cases, Workers, and Justice Institution
Privileged document leaking impacts the client, all staff in e-discovery, and the institution of justice.
1. Damages to client's cases, business and resources
Privilege document leaking is responsible for exposing confidential information, trade secrets, and privileged information. In the worst case, one single piece of leaked information can ruin the client case. Leaked damaging information can seriously injure the clients' future business, diminish its competitiveness, and invite chain lawsuits (when such information can be used by other parties in support of new suits). It is a huge waste for each of the great number of reviewers to repeat the same task.
2. Stress on all staff in e-discovery
When document review products are in such unpredictable quality, none of workers in e-discovery can control their liabilities. None of them can have a good sleep because they have to count on good luck. The odds of coding errors run litigation course in unpredictable way. If I look at those court reports, I can find that disputes in privilege document leaking are frequent. When more and more litigants know the potential leaking, they would try to find leaked documents in received documents, and so more and more disputes will appear in future court reports.
This creates a huge pressure on all workers in legal service delivery. Document reviewers, attorneys (associates and partners), and data service providers are subject to malpractice lawsuits. They can only hope that bad things will not happen, but bad things do happen according to cumulative probabilities. This is one of the main reasons for the extremely poor health condition of legal professionals. Each year, the Bar's study found that a high percentage of lawyers use caffeine, alcohol, pain killers, and sedative drugs.
3. Compromising the mission of delivering justice
When document review products are in such poor quality, adjudication is meaningless. Case disposition is often not based upon case merit. Case disposition depends upon conflicting, confusing, meaningless documentary evidence. Now, human review products are worse than work products of computers that are able to do several simple things with an overall IQ of one digit. This low quality problem is discussed frequently in court opinions. Delivery of justice has been seriously compromised by such e-discovery practice.
How to Address Privileged Document Leaking Problems
In any legal dispute, the party has only one or limited final objective. The most important point is avoiding mistakes in factual understanding. It is necessary to empower document reviewers and attorneys to know the basic information that is expected of intended readers of corporate documents.When documents are divided among many document reviewers, it is necessary to allow all reviewers to work as one person. It is necessary to use a review assisting system to help them share and seek information about basic facts, terms, and issues in the most efficient way. Due to the large volumes of basic facts, terms, and issues, one cannot use any conventional methods such as meeting, email, phone conference call, and black board posts etc. The system must provide real time or near real time data, and uses highly efficient search function.
I have worked on this problem since 2000, as the only person who has recognized and studied this problem. My solution can solve following problems: (1) Improving productivity by eliminating N-1 duplicate tasks; (2) using a method that allows document reviewers to code documents by guessing but also has the ability to cull those guessing-coded documents in a reconciling review; (3) reducing conflicting coding decisions as much as possible; (4) improving coding performance to the extend that the coding will not harm the client; and (5) providing the best capability of capturing sensitive information, trade secrets, and privileged information; (6) avoiding time on performing tasks that can only hurt client interests; (7) improving the team's ability to deal with routine changes; and (8) helping all workers and reviewers gain peace of mind.
This article discusses issues of general interest and does not give any specific legal or business advice pertaining to any specific circumstances. Before acting upon any of its information, you should obtain appropriate advice from a lawyer or other qualified professional.
This article may not be duplicated, altered, distributed, saved, incorporated into another document or website, or otherwise modified without the permission of TASA. Contact marketing@tasanet.com for any questions.