On December 12, 2016 the US Department of Justice (DOJ) Antitrust Division implemented an updated Model Second Request (Model), containing new guidance on the use of Technology Assisted Review (TAR) in responding to DOJ requests. The new guidance requires a party using TAR to provide the DOJ access to nonresponsive documents identified by the TAR algorithm. While courts have split over whether a party may be required to disclose nonresponsive documents used in TAR^[1], it is likely that this new guidance from the DOJ will significantly impact TAR-use negotiations in both civil litigation and regulatory forums.

The Model's instructions state that a party must submit a description of the search methods used. For any process that relies on TAR, a party must provide:

• "confirmation that subject-matter experts will be reviewing the seed set and training rounds;

• recall, precision, and confidence-level statistics (or an equivalent); and

• a validation process that allows for Department review of statistically-significant samples of documents categorized as non-responsive documents by the algorithm."

Adding further complication, the Federal Trade Commission (FTC) issued a revised Model Second Request in August 2015, which seeks the disclosure of TAR processes and analyses, but stops short of requiring nonresponsive documents. In contrast, the DOJ's revised Model requires access to nonresponsive documents, and tips the scales in favor of greater transparency in the use of TAR.

In-house counsel should anticipate that both civil litigants and regulators will cite this Model in their requests to access both nonresponsive documents used to train TAR algorithms and documents categorized as nonresponsive documents by a TAR algorithm. The disclosure of such information carries risks, such as the release of confidential proprietary information, and may open up the door for inquiries by other regulatory agencies or plaintiff's counsel in other actions. Ultimately, each party must weigh these risks against the immense e-discovery cost savings that TAR offers. But, as Hon. Andrew J. Peck recently noted, "[t]here may come a time when TAR is so widely used that it might be unreasonable for a party to decline to use TAR. We are not there yet."^[2]

_{[1] See Rio Tinto PLC v. Vale S.A., 306 F.R.D. 125, 129 (S.D.N.Y. 2015) (collecting cases, but declining to rule, on the need for the disclosure of non-responsive documents used to train TAR algorithms).}

_{[2] Hyles v. N.Y. City, No. 10CIV3119ATAJP, 2016 WL 4077114, at *3 (S.D.N.Y. Aug. 1, 2016).}