Privilege review, logging and clawback under FRE 502(d)

Privilege review does more than remove protected documents from a production. It must identify potentially privileged material, withhold it or redact only the protected portions, and describe each withheld item on a privilege log with enough precision for the other side to evaluate the claim without disclosing the substance of the protected communication.

That objective places privilege review at the intersection of legal judgment, collection design, document processing, and production control. A defensible workflow therefore cannot treat the privilege log as an administrative afterthought. Decisions about custodians, search terms, family relationships, redactions, and quality control determine both what reaches the production and what must be logged.

What privilege review must accomplish#

Privilege review has three connected outputs: identification, treatment, and explanation. Identification requires determining which documents appear to contain privileged material. Treatment means withholding an entire document, redacting protected portions while producing the remainder, or resolving the document through another agreed disposition.

Explanation means creating a log entry that states the asserted ground for withholding without revealing the protected content itself. The entry should allow the other side to assess whether the claim is facially supportable without learning the substance of the privileged communication. The information needed to accomplish that task depends on the format of the document and the agreement or order governing the log.

These outputs must remain consistent with one another. A document listed as withheld should correspond to a withheld item in the production set. A produced document containing a redaction should identify the basis for that redaction in the manner the parties have agreed. If an item is neither logged nor produced, the set requires investigation before release rather than after the receiving party identifies the discrepancy.

The review population also needs a defined beginning and end. Counsel may identify a candidate population through custodian selections, keyword searches, metadata filters, attorney-name screens, domain filters, or combinations of those methods. That population should then pass through attorney review so that automated candidates are evaluated and false positives are separated from materials that warrant protection.

Privilege review at volume#

The mechanics of a large review begin with candidate identification, but candidate generation is not a privilege determination. Counsel names and email domains can narrow a population efficiently, yet each method has predictable failure modes. A name screen may capture communications involving a person who did not hold a relevant role, while missing a communication whose metadata does not expose the expected identity.

Domain screens present a different problem. A domain may connect multiple people whose roles and responsibilities differ, so treating everyone associated with that domain as equivalent can over-collect and then inflate the review population. Conversely, a domain screen may miss material sent through another address, captured in a message header, or generated by a system that does not display the expected domain.

Name screens can also miss common variations in how a name appears. An automated screen should therefore be treated as a candidate-generation mechanism rather than the final classification mechanism. The review plan should account for the way names appear in the collection, including variations that may require reviewer attention.

Family-level review is equally important. An attachment is not automatically privileged because its parent email is privileged, and an email is not automatically nonprivileged because it is attached to a privileged document. Each attachment should be evaluated on its own content, context, and relationship to the rest of the family. Email threading and near-duplicate detection can help reviewers organize families, but they do not replace document-specific judgment.

De-duplication can reduce repetitive material before review, but the method should fit the requested production and the agreed log. If near-duplicates are collapsed, the resulting review population should preserve enough information to identify every document that must be logged. Near-duplicate detection may also identify materially different copies, so similarity should inform organization rather than serve as an unexamined conclusion.

Redaction requires a separate analytical step. When only part of a document is protected, the reviewer must distinguish the protected content from the material that can be produced. The redaction should correspond to the claim, and the log should describe the withheld information in a way that does not disclose that information. Producing a clean copy of a partially redacted document allows the receiving party to review the unprotected content without gaining access to the protected portion.

Quality control should address both sides of the production. The withheld set needs checks for complete log entries, correct document identifiers, consistent privilege designations, and the treatment of family members. The produced set needs checks for accidental privileged content, unexplained redactions, missing productions, and inconsistencies between the log and the delivered files.

A practical control process samples reviewer decisions and returns them for correction or second-level review. The sample should cover more than documents with identical outcomes. It should include keyword and name-screen candidates, complex families, partially privileged documents, large attachments, and items selected through de-duplication or near-duplicate detection.

Why FRE 502(d) changes the risk allocation#

Federal Rule of Evidence 502(d) permits a court to order that privilege or work-product protection is not waived by disclosure in the proceeding before the court. That mechanism makes disclosure-based clawback more defensible because the rule addresses the waiver question through a court order rather than leaving it solely to the parties’ agreement.

A party-to-party clawback agreement can define how disclosed material will be treated, how a receiving party must handle it, and what remedy applies if the agreement is breached. It does not replace privilege review. Materials that should never be produced must still be identified and withheld or redacted. The agreement provides a path for handling material that was disclosed mistakenly; it does not authorize disclosure.

Practitioners therefore distinguish between preventing disclosure and providing a remedy after disclosure. Review addresses the first. A clawback agreement and an FRE 502(d) order address the second by supplying a contractual or judicial mechanism for asserting that disclosure did not waive the protection in the relevant proceeding.

Seeking an FRE 502(d) order early gives counsel a court-based foundation before disclosure occurs. That foundation is particularly relevant when the review process prioritizes speed or permits broader initial disclosure of material that may be claimed as protected. The order does not make a privilege claim correct. It establishes that privilege or work-product protection may not be waived by disclosure in that proceeding, subject to the court’s order.

An order under FRE 502(d) should be obtained before material subject to a clawback agreement is disclosed. Otherwise, the agreement may still provide a contractual route to restoration, but the absence of a waiver finding does not undo the disclosure itself. The timing of the order can therefore determine whether the production strategy remains within a clear protection framework.

The rule should not be treated as permission to produce first and sort later. The same review controls remain necessary. The order may reduce the consequence of an erroneous disclosure; it does not eliminate the need to identify privileged material, log withheld items, or redact protected portions.

The logging burden#

A privilege log entry typically identifies the document and provides information about the author, recipient, date, subject or document type, and the asserted basis for withholding. It also describes the document without revealing the protected communication. The precise fields vary by document type, the sensitivity of the content, and the agreement between the parties.

For email, the log may need to distinguish the sender from recipients and addressees, identify copied recipients, and indicate whether the message was a request for legal advice, a communication conveying legal advice, or another category asserted by the producing party. For a document withheld as work product, the entry may describe the document and the asserted basis for the protection without disclosing its contents.

Metadata-only logging is often negotiated because a narrative description can itself reveal privileged information. Instead of describing the substance of a communication, the entry may present a limited set of metadata and a categorical designation. The parties must then agree on which fields are sufficient for the receiving party to assess the claim without providing the protected content.

Categorical logging follows a similar principle. Rather than describing each communication individually, the parties may agree to use broader descriptions for defined document categories, custodian groups, communication types, or privilege grounds. This can reduce repetitive narrative work, but it also affects the other side’s ability to evaluate the designation. The agreement should therefore state how categories apply and how they map to the documents identified in the log.

Document identifiers must work as the bridge between the log and the withheld set. Depending on the production design, that bridge may use a load file, document identifiers, family information, or another agreed reference. The format should preserve the ability to locate the exact withheld item and assess whether entries correspond to what was withheld.

Logging should also account for redactions. A produced document with a withheld portion may require a log entry that identifies the document, describes the basis for protection, and explains the redaction without reproducing the protected text. If the parties agree that redacted documents are described categorically, the production metadata should still make the affected items traceable.

Some review systems can generate portions of a log from document metadata. That does not eliminate attorney review because the attorney must confirm that the metadata accurately supports the asserted designation and that the entry does not disclose the protected content. Automated assembly can reduce clerical work, but it does not determine whether the claim belongs on the log.

The governing case-management instruction matters as well. Rule 26(f) requires the parties to confer before the scheduling conference, and Rule 16(b) governs the scheduling order. Those provisions do not prescribe a universal privilege-log format, but they provide the procedural structure through which the parties can address review, production, logging, and any agreed clawback process as part of their discovery planning.

The cost lever is scope, not reviewer speed#

The largest determinant of the privilege-review bill is the size and complexity of the population that reaches attorney review. Reviewer speed may affect staffing and elapsed time, but it cannot compensate for unnecessary collection, overbroad custodian definitions, disconnected name screens, or the failure to resolve document-family relationships before substantive review begins.

Scoping decisions therefore deserve attention before the review starts. The collection should be tied to the needs of the matter, the custodian population should be examined rather than treated as self-justifying, and search methods should be tested against known document characteristics. Counsel should also decide how attachments, duplicates, and near-duplicates will enter the workflow so that each category receives an appropriate review treatment.

Screening decisions have the same effect. A name screen may be useful for identifying candidates without defining privilege. A domain screen may narrow one part of the population while excluding another. The relevant question is what each screen contributes to candidate identification and what remains for attorney evaluation.

The production design also shapes the privilege burden. Load files, native files, image-based production formats, redactions, and family structures affect both review and logging. A workflow optimized for search may create a different logging burden than one organized around custodians or document types. Decisions should be made together rather than by review, processing, and logging teams in isolation.

FRE 502(d) provides a judicial protection framework, and a clawback agreement provides a contractual response to mistaken disclosure. Neither removes the need for careful review. The defensible and cost-conscious approach begins with the collection and screening decisions that determine who must review what, then aligns privilege analysis, redaction, log design, and production controls with that population.