skip to content

Case Studies in Privilege Review: Northrop Grumman Sys. Corp. v. United States

by Joshua Gilliland


Privilege review requires planning. Attorneys need to ask: what privileges are at issue in a lawsuit? Who could possibly be claiming a privilege? What are the elements of that privilege?

Many parties today include a clawback agreement into their discovery orders. Such clawback agreements require, under the Federal Rules of Evidence Rule 502(b)(2) and (3), that the holder of the privilege conducts reasonable document review and in the event of an inadvertent disclosure, that the assertion of the privilege is reasonably prompt.

Northrop Grumman Sys. Corp. v. United States is a case study of how privileged documents can be inadvertently produced. Fortunately for the Defendants in the case, they were found to have not waived their privilege—despite the specter of an unreasonable privilege review and not-prompt assertion of privilege. (Northrop Grumman Sys. Corp. v. United States, 2015 U.S. Claims LEXIS 235, 6 (Fed. Cl. Mar. 9, 2015).)

Here is what happened: the data at issue was on two hard drives that were delivered to a document review service provider. The service provider was supposed to conduct a search for the terms “privilege,” “work product,” or the government attorneys whose names appeared with the subject contract anywhere in the document. (Northrop, at *6.

As one can guess, this was an overbroad search strategy that identified 560,000 documents. This should be no surprise, considering that many professionals who are not attorneys have “confidential attorney work product” messages in the footer of their emails. Such business practices create a high volume of “false positive” search results. Moreover, there could be a strong argument that those who claim everything is an attorney-client communication could end up waiving the privilege.

The document review service provider narrowed the privilege review search to documents only sent to or received by attorneys. This search resulted in 83,000 “hits.”

Production followed on October 31, 2013. A chronology of events began on November 26, 2013, that alerted the Defendant to inadvertently-produced documents that resulted in the clawback provision being invoked once in December of 2013 and again in June of 2014 for 1,500 privileged files. (Northrop, at *6-7.)

Defense Counsel admitted to making two mistakes with the service provider.  First, the attorney was “under the impression that the third-party vendor was searching, not just for attorney names, but for terms such as “privileged” or “work product.” This was not the case. Second, Defense Counsel was “under the impression that the ‘to and from’ attorney search among emails included such addresses anywhere in the chain of emails included in the document, not just the last email.” This, too, was incorrect. (Northrop, at *7-8.)

The Court stated that Defendant’s error was not so careless to amount to a waiver. Furthermore, the Court declined to rule that there had been unreasonable delay, given the number of letters sent between counsel. However, the Court did state that “[p]lainly there was too much reliance on the third-party vendor. Greater care and oversight should have been exercised by counsel.”

How can attorneys exercise greater care in privilege review? One way is to leverage the technology of Everlaw’s Prediction Engine. Consider what goes into the attorney-client privilege: there has to be a communication from a client to a lawyer seeking legal advice, or a communication from a lawyer giving legal advice to a client.

With predictive coding, legal professionals can easily create a prediction model around privilege classifications.

A reviewing attorney can begin their document review by searching for emails from the attorney to the client. Once identified, the reviewing attorney can code those files as falling within the attorney-client privilege, or not. All the while, the Prediction Engine is continuously learning from how documents are classified, thus enabling speedy identification of other potentially privileged documents—simply run a search for all documents predicted to be privileged. The Prediction Engine also facilitates reliable QC of privilege review—search for all documents likely to be privileged, but classified as not privileged, or vice versa.

The same workflow can be employed for reviewing trade secrets, HIPAA, or other confidential information. Additionally, privileged communications such as the spousal privilege or the clergy privilege could be identified and categorized in a similar fashion. This does require those setting up document review to outline and create coding protocols for the possible privileged communications in the case or sources of confidential information. However, this time is well spent, as careful planning of your privilege review strategy can help protect confidential information from accidental disclosure.

No lawyer wants to explain to a client how a large volume of privileged information was disclosed to the opposing party. Predictive coding is one means of ensuring your privilege review is a reasonable means of protecting confidential information. It is a superior form of review compared to searching for “work product” and getting every email with a confidentiality statement in the footer. Moreover, few people begin a privilege communication by writing the word “Confidential” in the email subject line. What matters is the context of the communications. Everlaw’s Prediction Engine can learn from the nuance of actual attorney review to help expedite identifying what is potentially privileged information in discovery.