aiR for Privilege
aiR for Privilege is an AI-powered solution that uses generative AI, large language models, natural language processing, machine learning, and social network graphs to analyze documents and predict whether they are privileged or not and provide a draft privilege log description. This streamlines privilege review, improves the accuracy of your final privilege calls, and reduces overhead when creating a privilege log.
aiR for Privilege retains an understanding of privilege specific to an organization to drive greater consistency in privilege calls across matters and speed up the pace of privilege review.
Note: aiR for Privilege locates attorney-client privileged communications and documents subject to the attorney work-product doctrine, as understood under US standards and laws. It is up to the user of aiR for Privilege to decide whether this criteria and the results meet applicable requirements.
Note: Azure OpenAI does not retain any data from the documents being analyzed. Data you submit for processing by Azure OpenAI is not retained beyond your organization’s instance, nor is it used to train any other generative AI models from Relativity, Microsoft, or any other third party. For more information, see the white paper A Focus on Security and Privacy in Relativity’s Approach to Generative AI.
Note: For European Economic Area (EEA) customers, aiR for Privilege data may be processed elsewhere in the EU, but it will always be processed in compliance with applicable laws. For more information, please contact your account manager.
Conceptual workflow
The table below illustrates the high-level workflow and the role that should be performing each step for aiR for Privilege:
| Role | Step | Workflow Step |
|---|---|---|
| Project Manager | 1 | Create a saved search of documents that you want to analyze for privilege |
| 2 | Create an aiR for Privilege project | |
| 3 | Upload any lists of Known Attorneys and Known Law Firms to jump-start the AI’s understanding | |
| 4 | Run each aiR for Privilege pipeline step | |
| Annotator | 5 | Confirm AI Predictions through Annotations during each pipeline step |
| Case SME | 6 | Finalize which Law Firms and Attorneys are Privilege Conferring, Privilege Neutral, or Privilege Breaking |
| Project Manager | 7 | Generate Privilege Predictions and Draft Privilege Log Descriptions |
| Case SME | 8 | Finalize privilege decisions and log descriptions through targeted quality check (QC) |
| 9 | Produce with confidence |
For more information about the different aiR for Privilege roles see Roles and security permissions.
Requirements
Following are requirements for using aiR for Privilege:
- You must complete aiR for Privilege contracting to use it in your RelativityOne instances.
See Billing for more details. Contact your Relativity Sales Representative for contracting. - Relativity must install the aiR for Privilege application to your instance.
- You must install the aiR for Privilege application to your workspace.
See Installation and configuration for more details. - You should install the Import/Export application to your workspace as it can be helpful in accelerating the addition of Known Items and performing the Confirm Privilege Status step.
Considerations
Following are limitations to consider when using aiR for Privilege:
- aiR for Privilege is available in RelativityOne instances hosted in Central US, East US, and UK South Azure data centers. Support in other data centers coming soon.
- There are limitations on both what types of documents and the number of documents that can be included in an aiR for Privilege project as specified in Saved search requirements.
- Only a single aiR for Privilege project within a workspace can be In Progress at any given time.
- We tested the performance of aiR for Privilege on primarily English language documents. Use of aiR for Privilege with non-English language documents may result in unexpected or unverifiable results. Consider using Language Identification to better understand the language distribution across your data set.
Note: aiR for Privilege only analyzes emails that have email headers in the following languages: English, Chinese, Dutch, French, German, Japanese, Korean, Portuguese, Spanish.
- aiR for Privilege locates attorney-client privileged communications and documents subject to the attorney work-product doctrine, as understood under US standards and laws.
-
aiR for Privilege supports analyzing emails, attachments, and standalone documents. It does not currently support the analysis of RSMF documents.Note: RSMF documents that are included in a project will by default be analyzed as standalone documents. If the extracted text for an RSMF document contains valid email headers (from, to, date, subject, etc.) and has a Record Type of email it will be analyzed as if it is an email with it's senders and recipients considered.
- In progress aiR for Privilege projects will be abandoned when the workspace is deleted, regardless of Recycling Bin, archived (ARM), or put in Cold Storage.
- aiR for Privilege only works in Review workspaces.
- aiR for Privilege does not currently support CMK instances.
- aiR for Privilege is not offered in RelativityGov or Relativity Server.
Billing
aiR for Privilege is a usage-based product, meaning you pay-per-use. Charges are per Document Unit. A Document Unit is equal to one Relativity document.
After clicking the Run button on the Prepare Project pipeline step, Relativity estimates the number of Document Units you’ll use for this project. You can then decide to continue or cancel the step.
After completing the Prepare Project pipeline step Relativity will charge you for the number of Document Units. The price per Document Unit may vary. Your RelativityOne contract defines the price.
Note: You may have included aiR for Privilege Document Units to trial the product without any cost. After you use your Included aiR for Privilege Document Units, Relativity will charge per Document Unit.
Brain
The Brain stores historical annotation decisions that can then be applied to an existing project. This accelerates annotation time and ensures consistency across the predictions that the AI makes. Annotation objects found in the Brain will have the Source field populated accordingly.
The Brain contains three levels.
- The Universal Brain, which contains items that are public knowledge that will be applied to every project.
- The Client Brain, which contains client-specific items that can be applied to different matters for the same client.
- The Matter Brain, which contains workspace-specific items that can be applied to different projects for the same matter.
Universal Brain
The Universal Brain stores common, publicly available, knowledge that helps drive more exact and consistent privilege predictions. Disclaimer calls, Public, 3rd Party and Law Firm Domain calls, and Equivalent Domain calls are stored in the Universal Brain.Customer data is not extracted and stored in the Universal Brain.
Client Brain
After the completion of a project, aiR for Privilege will extract and store select information from that project in a Client Brain that is specific to the Client object associated with the workspace. Any future project in a workspace associated with that Client object will have the opportunity to use knowledge stored in the Client Brain to jump start the decision-making process, resulting in greater consistency. See General settings in a workspace for more details on using the Client Brain. In-House Domain calls and Entity calls are currently stored in the Client Brain.
Note: We plan to add Disclaimers, Domains, and Equivalent Domains to the Client Brain for general availability.
Matter Brain
The Matter Brain is data stored within the Disclaimers (Priv), Spammers (Priv), Domains (Priv), Equivalent Domains (Priv), Entity (Priv) tabs within a single workspace used for later projects within that workspace. When a new project locates an artifact that was already discovered it will link the new project to that artifact and take that artifact’s Annotator Decision as truth. This drives consistency and accelerates annotation time.
Regional availability of aiR for Privilege
The availability of aiR for Privilege varies by region. The availability of the large language model (LLM) used by aiR for Privilege also varies by region.
The following table shows the current LLM model and when aiR for Privilege will be available for each region:
|
Region |
Current LLM Model |
Date aiR for Privilege is Available |
|---|---|---|
|
United States |
GPT-4 Omni |
2024-11-1 |
|
United Kingdom |
GPT-4 Omni |
2024-11-1 |
|
Canada |
GPT-4 Omni |
2024-11-1 |
|
Australia |
GPT-4 Omni |
2024-11-15 |
|
France |
GPT-4 Omni |
2024-11-15 |
|
Germany |
GPT-4 Omni |
2024-11-15 |
|
Ireland |
GPT-4 Omni |
2024-11-30 |
|
Netherlands |
GPT-4 Omni |
2024-11-30 |
|
Switzerland |
GPT-4 Omni |
2024-11-30 |
