Privilege 2.0

Privilege 2.0

Privilege 2.0 Guide

***The New Privilege Application was made available April, 2026.  If you are using an older plug-in, please go to the other legacy version of Privilege Review Help for assistance.

eDiscovery AI Privilege Review

Privilege review is one of the most time intensive and important aspects of document review.  Identifying and withholding or redacting privileged documents usually requires a substantial review team, scrutinizing tens or hundreds of thousands of documents, usually hitting on a very broad search term screen of various attorney names and terms implying potential privilege. eDiscovery AI’s privilege application can review, identify privilege and draft a privilege log entry based on your instructions and customized input.  

 

Before Submitting Documents

Create separate fields for the following items, all of these fields are required to submit documents for Privilege review.   

(1) Privilege Classification (Single Choice)

(2) Privilege Type (Multi Choice)

(3) Attorney List (Long Text)

(4) Privilege Explanation (Long Text) 

(5) Privilege Log Entry Field (Long Text) 

Select the documents eligible for privilege review and use the mass action Send to eDiscovery AI

 

 

 

 

From the Select Job Type list, click Privilege, then click Next.



Note, if you do not see a popup window double check that any popup blocking feature on your browser is set to allow popups from your Relativity instance.  

 

What follows below is a step by step of each of the individual inputs used for Privilege Review.

 

Load Settings

The Load Settings button will allow you to automatically populate the settings, prompts, and results fields used from previous jobs. Click the Select Previous Job drop down to see a list of previous jobs submitted. The “Review PDFs as Images” option won’t carry over—you’ll need to check it manually. Custom Privilege Types are not currently carried over from previous jobs and must be re-added manually. 

 

Review PDFs as Images

eDiscovery AI can process both extracted text documents and image documents. For PDFs, by default the extracted text is used. However, you have the option to pull the information eDAI analyzes directly off the image itself. This is particularly beneficial in various situations:

·    Scanned documents, particularly if the OCR extracted text isn’t very good or non-existent.

·       Script handwriting analysis, or handwritten notes.

·       Landscaped documents where OCR regularly fails.

 

General instructions

This is where one can enter the general directions as to how they want Privilege handled for classification. Instructions on how narrow or broad to treat it, whether footers should be included or ignored, and whether various targeted documents should be included can be added here.


Privilege Types

Privilege Types refers to what types of Privilege you would like eDiscovery AI to consider in its review and the definitions used to classify those types.  There are two default options available, which only need to be checked if you wish for eDiscovery AI to identify and tag them accordingly:

 

Attorney Client Privilege

 

Attorney/Client will apply the eDiscovery AI Attorney Client Communication Privilege prompt to the documents. Above is the default, generalized description of Attorney-Client Privilege. If you would prefer a more customized definition of Attorney Client privilege to be applied, you are free to edit the definition as you deem necessary.     

 

Work Product


 Work Product will apply the eDiscovery AI Work Product prompt to the documents. Above is the default, generalized description of Work Product. If you would prefer a more customized definition of Work Product be applied, you are free to edit the definition as you deem necessary. 


Additional Privilege Types

One can create additional Privilege Types on a submission basis by clicking the “+Add Privilege Type” button in the lower right side of the Privilege Types section.  This will add another entry area for you to Name and provide a Description for any additional Privilege Types.

 

Attorney List

Attorney List 
Please enter a list of attorneys or law firms which will help eDiscovery AI identify and appropriately label documents as privileged.  The more comprehensive the list is, the better the system will be at identifying privileged materials. No special formatting or terms and connectors are needed for this list. Please note that if you use generic terms such as “attorney” or “general counsel” on the privilege list, then people associated with those terms within the documents may be listed on the Attorney List Field output even if not explicitly included (e.g., “John Doe, Attorney” may result with “John Doe” being listed on the Attorney output even though he is not listed).

 

Attorney List Instructions

 

Attorney List Instructions
This is where the instructions of how one would like the individual attorney names output to the Attorney Field can be addressed. One can specify a limit on attorney names per document, whether new attorneys found (not named in the Attorney List) should have an asterisk next to their names, and other ways to format names in this field.

 


Log Configuration And Picklist

 

 

 

The Log Configuration section is where one enters instructions of how you would like your Privilege Log Entry constructed. Various rules, examples and specified texts can be included directly to this section in an organized format.  These rules can be used separately or in conjunction with picklists.  If a picklist is used, any specific rules for various options can still be included to the respective Log Configuration rule, but the Picklist will contain the various options available. 

 

One can create up to six different picklists to choose from in supplementing the Log Configuration.  References to Picklists in the Log Configuration should specify the NAME of the Picklist (not the Picklist Number). Within the "Options" section, all possible options you would like for that particular section of the Privilege Log Entry should be included, with a semicolon at the end of each option, and only one option per line. Any guidance for which option is appropriate can be maintained within the Log Configuration section. 

·       Best Practice Tip:
      Capitalize the entire name of the Picklist.  Then when you reference it in the Log Configuration section, capitalize it there as well. It will make the organization easier to follow for both you and the LLMs. 

If an option from the picklist is required for the privilege log, select the “Required” toggle. If the options for the picklist are more freeform, where you don’t have specified options and just want the AI to create the specific content, select “Freeform.”

Result Mapping

 

 

This is where one would specify which fields the results of your eDiscovery AI Privilege Review run should be populated. All of these (with the exception of the Family Identifier Field) should have been created prior to the “Before Submitting Documents” section of the instructions.

a.   Family Identifier Field.  The field the AI Classifier can use to identify family groupings.

b.   Privilege Classification. The field identifying whether a document was found to be privileged or not privileged accordingly.
c.   Privilege Type. The type(s) of privilege identified within the document, as determined by your descriptions for the various privilege types.
d.   Attorney List. The list of attorneys found within the document in the format described in the “Attorney List Instructions” section.
e.   Privilege Log. The Privilege Log entry as defined by the instructions within the Log Configuration and Picklist sections.
f.    Privilege Explanation. An AI generated description as to why the document was identified as privileged given the instructions.

 

Submitting Documents and Viewing Results

Once all the forementioned sections are entered, you can click “Next.”  That will take you to a confirmation page, reiterating all the content and rules that were entered.  If all looks correct, click “Submit.”  You should get confirmation the job was submitted.

·      Best Practices Tip: It is always a good idea to Print this page to PDF to maintain a record of the submission for future reference.  That button can be found at the top right of the page. [Printer icon]

Once eDiscovery AI completes its analysis and populates the fields identified, you should see and be able to use the results in the same way as any other field in Relativity such as searching and exporting for a privilege log.  

 

·       
Best Practices Tip:
  It is a good idea to incorporate these fields into a view or a layout for which you will be considering and QCing the results. 

 

 

Important Considerations 

While eDiscovery AI is incredibly effective, it’s still recommended that attorneys conduct QC on the results. Some QC steps to consider:  

·     Any document lacking any respective privilege elements should be considered for QC and manual review.  It is likely that some of these may require redaction or are no longer privileged; that decision should come from the attorney, not the AI model.  

·       A percentage of documents hitting on various terms or names should be considered for sampling and review. If something is consistently categorized incorrectly, you may want to search on all such documents or revise prompt instructions.  

·       Priv Log Entries should be reviewed for grammar, context and quality, and edits made accordingly.  

·       If results are consistently overinclusive or underinclusive, please advise your provider and the possibility of either narrowing or broadening the prompts respectively. 

 

Privilege checks are always an important part of any review and production process.  It is recommended that conducting an eDiscovery AI sweep across your potential production set to ensure no privileged documents are disclosed is a sound best practice and good QC check for your production processes.    


    • Related Articles

    • eDiscovery AI Privilege Review

      1) Privilege review is one of the most time intensive and important aspects of document review. Identifying and withholding or redacting privileged documents usually requires a substantial review team, scrutinizing tens or hundreds of thousands of ...