On September 17, the U.S. Tax Court in Dynamo Holdings Ltd. P’ship v. IRS, 143 T.C. No. 9 added itself to the growing list of courts that have approved the use of predictive coding in litigation.

As we have previously noted, predictive coding or Technology Assisted Review (“TAR”) has increasingly been utilized in large scale document productions in a wide variety of litigation and government investigation matters. However, not all parties and authorities have embraced the use of TAR, perhaps due to litigation’s adversarial nature, or a latent fear that technological tools will somehow miss key documents that a manual document-by-document review would otherwise catch. As the body of cases and research articles grows, however, courts and academics have largely rejected these concerns in favor of the more efficient, less expensive, and, arguably, more accurate document discovery that predictive coding offers. Yet in many jurisdictions, the use of predictive coding in litigation still remains untested.

Enter the U.S. Tax Court. In Dynamo Holdings, the issue was whether certain transfers from one business entity to another (the “Petitioners”) constituted gifts or loans. The IRS requested that the Petitioners produce electronically stored information contained on two backup storage tapes, which both sides conceded likely had relevant information.

Petitioners resisted a linear review of the backup tapes, which they estimated would “take many months and cost at least $450,000.” Petitioners also resisted the IRS’s offer to produce the backup tapes without review, with the opportunity to clawback any privileged materials discovered (especially since the tapes purportedly contained sensitive, and otherwise confidential documents, including “personal identification information … [and] HIPAA protected information” that was irrelevant to the requests). Instead, Petitioners proposed that they produce responsive documents identified by predictive coding.

The Tax Court held an evidentiary hearing about whether predictive coding should be used and heard expert testimony from both parties. Petitioners’ expert estimated that predictive coding would reduce the cost of document review and production for Petitioners by more than 80%. That expert also noted that predictive coding offered efficiencies over keyword searching and the ability to test and verify the inclusiveness of a production through the “use of statistics and other sampling information.”

The IRS’s expert “did not persuasively say anything to erode or otherwise undercut” the Petitioner’s expert, leaving the IRS’s only argument that predictive coding constituted an “unproven technology.”

The Tax Court ultimately sided with Petitioners, concluding that predictive coding provided a “potential happy medium” between the IRS’s need for requested documents and Petitioners’ desire to decrease the burden of review and production. In so doing, the Tax Court found that predictive coding furthered the Tax Court’s Rules, which call for “secur[ing] the just, speedy, and inexpensive determination of every case.” In the event that the IRS believed that Petitioners’ production of documents failed to satisfy its requests, the Tax Court noted that the IRS could file a motion to compel.

In reaching its decision, the Tax Court became one of the first Article I courts to accept predictive coding as an approved discovery tool. It remains to be seen whether other specialty courts or administrative agencies will follow suit.