Synthetic datasets

Each document thread contains a predefined “nugget”. This nugget is the document’s primary piece of information. It determines the privilege status, either explicitly or implicitly. All nuggets are predefined before generation, and documents are built around nuggets to ensure control over privileged and non-privileged content. Note: If a privileged nugget is included in a document with a third party cc, the document is not privileged

Privileged Scenarios

These scenarios represent communications that can be tagged as “Attorney Client Privilege” and/or “Work Product Doctrine”. 1. Litigation Strategy Advice An attorney is providing strategic guidance specifically related to pending or anticipated litigation, to the client. This is protected by BOTH attorney-client privilege AND work product doctrine. Key characteristics:

Attorney is actively involved
Discussion involves litigation planning or strategy
Contains attorney’s mental impressions and case analysis
No third parties are copied or involved

Nugget topic distribution: 3 variations per topic

Nugget Category	Number of Topics	Example
litigation_strategy	4	”Attorney providing opinion on settlement versus litigation decision”

2. Attorney Providing Advice A lawyer is giving professional legal advice to a client. The communication is confidential, with no third parties present. This is protected by attorney client privilege. Key characteristics:

Attorney provides legal analysis or recommendations
Communication is private (no third-party CCs)
Purpose is to provide legal advice

Nugget topic distribution: 59 total nugget topics (3 variations per topic)

Nugget Category	Topics	Example
contract_review	6	”Attorney reviewing and analyzing contract terms, obligations, and risk allocation”
employment_matters	6	”Attorney advising on employment law, workforce decisions, and workplace compliance”
litigation_strategy	6	”Attorney providing guidance on dispute resolution, litigation management, and case strategy”
general_risk_assessment	6	”Attorney evaluating legal risks and liability exposure for business decisions”
corporate_governance	6	”Attorney advising on board responsibilities, shareholder rights, and governance structures”
IP protection (tech industry)	1	”Attorney advising on intellectual property strategy for technology innovations”
Data privacy compliance (tech)	1	”Attorney advising on data protection laws and privacy obligations”
Clinical trial compliance (healthcare)	1	”Attorney advising on FDA regulations for clinical trials”
Securities compliance (finance)	1	”Attorney advising on SEC disclosure requirements”
… 25 more categories	…	…

3. Client Seeking Advice What it means: A client (employee or executive) is asking their attorney a legal question or requesting legal guidance. This falls under attorney client privilege. Thi Key characteristics:

The client is inquiring about a certain topic and will explicitly OR implicitly ask for advice.
Purpose is to obtain legal guidance
No third parties are present

Example topics:

Asking attorney to interpret contract terms
Seeking guidance on employment decisions
Requesting review of proposed business transactions
Asking about compliance obligations
Seeking advice on IP matters or data incidents

Nugget topic distribution: 57 nuggets (3 variations per topic)

Category	Topics	Example
contract_interpretation	1	”Client asking attorney to interpret contract terms or assess contractual rights”
employment_questions	1	”Client seeking attorney guidance on employment decisions and workplace matters”
transaction_approval	1	”Client seeking attorney review and approval for business transactions”
crisis_response	1	”Client seeking immediate attorney guidance on urgent legal crisis”
compliance_interpretation	1	”Client asking attorney to clarify regulatory compliance requirements”
IP questions (tech industry)	1	”Client seeking attorney advice on intellectual property matters”
Data incident response (tech)	1	”Client seeking attorney guidance on data security incidents”
Patient privacy (healthcare)	1	”Client seeking attorney advice on HIPAA compliance”
… 21 more categories	1 each	…

4. Attorney Directs Document Collection What it means: An attorney is directing an employee to collect, preserve, or organize documents specifically in anticipation of litigation or for litigation case preparation. This is protected by Work Product Doctrine but not by Attorney Client Privilege. Key characteristics:

Attorney is explicitly directing the activity
Purpose is litigation preparation
Attorney is orchestrating fact-gathering or document preservation
Collection is at attorney’s specific direction

Example topics:

Directing document preservation for anticipated litigation
Requesting specific fact-gathering for case preparation
Asking employee to compile documents under attorney’s guidance for litigation

Nugget topic distribution: 4 Topics (3 variations of each)

Category	Topics	Example
document_collection	4	”Attorney directing employee to collect and preserve documents for anticipated litigation”

Non-Privileged Scenarios

These scenarios represent communications that are NOT protected by privilege, even though they may sometimes look similar to privileged communications. 5. Business Discussion What it means: Regular employee-to-employee business communications about day-to-day work topics. Example topics:

Project planning and execution
Product development discussions
Sales and customer relationship management
Operations and process improvements
Technical architecture decisions
Marketing campaigns

Nugget topic distribution: 28 categories, 555 nugget topics (3 variations per topic)

Category	Topics	Example
marketing_strategy	90	Business marketing and brand positioning discussions
project_coordination	86	Cross-functional project management and execution
financial_planning	80	Budget planning, revenue analysis, financial metrics
operational_planning	80	Operations strategy, process improvements, efficiency
strategic_initiatives	80	Business strategy, growth planning, market expansion
product_development	29	Product roadmap, feature planning, development cycles
… 22 more categories	…	Clinical operations, regulatory strategy, quality assurance, manufacturing, etc.

6. Attorney Non-Legal An attorney is involved in the conversation, but the discussion is about non-legal business matters. The attorney is acting in a business capacity, not providing legal advice. Key characteristics:

Attorney is a participant
Topic is business-related, not legal
Examples: budgeting, vendor selection, office management, business strategy
Attorney is acting as a business person, not legal counsel

Nugget topic distribution: 1 nugget type (3 variations)

Topic Category	Count	Example
administrative_coordination	1	”Legal team coordinating on administrative processes”

7. Attorney Providing Advice WITH Third-Party CC What it means: An attorney is providing legal advice, but a third party (someone outside the attorney-client relationship) is copied on the communication. Nugget topic distribution: 34 categories, 59 total nugget topics (same as Attorney Providing Advice)

8. Client Seeking Advice WITH Third-Party CC What it means: A client is seeking legal advice from their attorney, but a third party is copied on the request. Key characteristics:

Client seeks legal advice
Third party is CC’d
Third party could be external consultant, vendor, personal contacts, etc.
Same intent as “Client Seeking Advice” but not protected

Nugget topic distribution: 29 categories, 57 nuggets (same as Client Seeking Advice)

9. Privilege Mimicking What it means: A non-lawyer employee from a legal firm uses legal-sounding language or discusses legal topics, creating the appearance of attorney-client privilege when none exists. Key characteristics:

No attorney involved (may be paralegal, legal assistant, compliance officer, etc.)
Uses legal terminology or discusses legal concepts
Creates false appearance of privileged communication
Purpose is to confuse privilege detection systems

Nugget topic distribution: 1 nugget topic (3 variations)

Category	Count	Example
legal_terminology_mimicking	1	”Non-lawyer referencing compliance policies using formal legal terminology”

10. Work Product Mimicking What it means: A non-lawyer from a legal firm is directing document collection or conducting investigations that look similar to attorney work product, but are for business purposes, not litigation. Key characteristics:

No attorney directing the activity
Purpose is business/operational (compliance audits, due diligence, etc.)
May involve document collection, investigations, or fact-gathering
Looks like work product but lacks litigation purpose

Example topics:

Routine compliance audits (SOC 2, ISO certifications)
Due diligence for M&A or vendor selection
Operational efficiency studies
HR investigations for workplace culture
Security assessments

Nugget topic distribution: 5 categories, 15 total topics (3 variations per topic)

Category	Topics	Example
compliance_audits	3	”Non-lawyer requesting document compilation for routine SOC 2 or ISO 27001 compliance audit”
business_document_collection	3	”Non-lawyer requesting document gathering for M&A due diligence or vendor selection process”
internal_investigations	3	”HR or operations requesting document collection for workplace culture assessment”
regulatory_filings	3	”Non-lawyer coordinating document assembly for periodic regulatory reporting”
records_management	3	”IT or operations requesting document retention policy implementation”

11. Non-Lawyer Business What it means: Communications between company employees and law firm staff (paralegals, legal assistants, legal secretaries) about business or administrative matters, not legal advice. Key characteristics:

Law firm staff (non-attorneys) communicating with company
Topics are administrative, logistical, or business-related
No legal advice or strategy discussion

Nugget topic distribution: 1 nugget type

Category	Category	Example
legal_technology_discussion	1	”Discussing legal technology tools and system implementations”

12. Company-Third Party What it means: Communications between company employees and external third parties such as vendors, customers, consultants, auditors, or partners. These are standard business relationships. Key characteristics:

External third party involved
Business purpose

Example topics:

Vendor negotiations and contract discussions
Customer support and issue resolution
Consultant project management
Auditor compliance reviews
Partner coordination and collaboration

Nugget topic distribution: 5 categories, 22 total topics (3 variations per topic)

Category	Topics	Example
vendor_negotiation	5	”Company employee negotiating SaaS contract pricing and license tiers with vendor”
customer_support	5	”Company employee resolving customer technical issues and escalated complaints”
consultant_engagement	4	”Company employee defining consultant project scope and success criteria”
auditor_compliance	4	”External auditor requesting financial records and supporting documentation”
partner_coordination	4	”Company employee coordinating integration roadmap with strategic technology partner”

13. Representing to Opposing Counsel What it means: Communications between attorneys representing opposite sides in a litigation matter. These are adversarial communications during active litigation. Key characteristics:

Attorneys from opposing sides communicating
Talking about active litigation case
Topics include discovery, settlement, or procedural matters

Example topics:

Discovery coordination and ESI protocols
Settlement negotiation and offers
Procedural scheduling and court coordination
Document production logistics

Nugget topic distribution: 3 categories, 10 total nugget types

Topic Category	Count	Example
discovery_coordination	4	”Negotiating ESI search terms to reduce over-collection and production costs”
settlement_negotiation	3	”Proposing settlement terms and financial resolution framework”
procedural_scheduling	3	”Requesting extension of motion filing deadlines with professional courtesy”

Communication Flows

Communication flows define the multi-message sequences in an email thread. They control who sends what to whom and in what order. This allows us to model the following threads:

Direct

A single message from a primary to a secondary. No CC, no reply.

Reply

Primary sends the initial message with the nugget. Secondary replies in the same thread

With CC

Single message with an internal peer CC’d

With Outsider CC

Single message with a third party CC

Forward

Primary sends message to secondary who forwards message to a peer

Escalation

A four-step flow where the secondary escalates to their manager, the manager responds, and the secondary relays back to the primary.

Attachment

Single message that also includes an attachment in the email

Document Review

Primary sends a document for a review. Secondary replies with feeback, optionally attaching a revised version.

Overview

Fundamentals

AI Chat

Work with projects

Release Notes

Product Guides

System

Use Ask AI

Change logs

Sandbox

Troubleshooting

Internal

Epiq AI V2

Synthetic datasets

Privileged Scenarios

Non-Privileged Scenarios

Communication Flows

Direct

Reply

With CC

With Outsider CC

Forward

Escalation

Attachment

Document Review

​Privileged Scenarios

​Non-Privileged Scenarios

​Communication Flows

​Direct

​Reply

​With CC

​With Outsider CC

​Forward

​Escalation

​Attachment

​Document Review

Privileged Scenarios

Non-Privileged Scenarios

Communication Flows

Direct

Reply

With CC

With Outsider CC

Forward

Escalation

Attachment

Document Review