Privilege Detection Overview

The Problem

E-discovery teams review thousands of documents manually to identify privileged content. Miss one privileged document in production, and you’ve waived protection. Over-designate, and you face sanctions for obstruction.

The Solution

An AI-powered privilege detection API that analyzes documents for attorney-client privilege, work product doctrine, and other protected communications. Get confidence scores, category breakdowns, and actionable recommendations for each document.

Quick start

# Analyze content for privilege
curl -X POST https://api.case.dev/privilege/v1/detect \
  -H "Authorization: Bearer $CASEDEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Attorney memo: Confidential legal advice regarding...",
    "categories": ["attorney_client", "work_product"],
    "include_rationale": true
  }'

casedev privilege:v1 detect \
  --content "Attorney memo: Confidential legal advice regarding..." \
  --category attorney_client,work_product \
  --include-rationale

import Casedev from 'casedev';

const client = new Casedev({ apiKey: process.env.CASEDEV_API_KEY });

// Analyze document content for privilege
const result = await client.privilege.detect({
  content: 'Attorney memo: Confidential legal advice regarding...',
  categories: ['attorney_client', 'work_product'],
  include_rationale: true,
});

if (result.privileged) {
  console.log(`Privileged (${result.confidence * 100}% confidence)`);
  console.log(`Recommendation: ${result.recommendation}`);
  
  // Review detected categories
  for (const cat of result.categories) {
    if (cat.detected) {
      console.log(`  ${cat.type}: ${cat.rationale}`);
    }
  }
} else {
  console.log('No privilege detected - safe to produce');
}

// Or analyze a document already in your Vault
const vaultResult = await client.privilege.detect({
  document_id: 'doc_abc123',
  vault_id: 'vault_xyz789',
});

import casedev

client = casedev.Casedev(api_key=os.environ['CASEDEV_API_KEY'])

# Analyze document content for privilege
result = client.privilege.v1.detect(
    content='Attorney memo: Confidential legal advice regarding...',
    categories=['attorney_client', 'work_product'],
    include_rationale=True,
)

if result.privileged:
    print(f'Privileged ({result.confidence * 100}% confidence)')
    print(f'Recommendation: {result.recommendation}')
    
    # Review detected categories
    for cat in result.categories:
        if cat.detected:
            print(f'  {cat.type}: {cat.rationale}')
else:
    print('No privilege detected - safe to produce')

# Or analyze a document already in your Vault
vault_result = client.privilege.v1.detect(
    document_id='doc_abc123',
    vault_id='vault_xyz789',
)

result, _ := client.Privilege.V1.Detect(ctx, casedev.PrivilegeV1DetectParams{
	Content:          casedev.F("Attorney memo: Confidential legal advice regarding..."),
	Categories:       casedev.F([]string{"attorney_client", "work_product"}),
	IncludeRationale: casedev.F(true),
})

if result.Privileged {
	fmt.Printf("Privileged (%0.f%% confidence)\n", result.Confidence*100)
	fmt.Printf("Recommendation: %s\n", result.Recommendation)

	for _, cat := range result.Categories {
		if cat.Detected {
			fmt.Printf("  %s: %s\n", cat.Type, cat.Rationale)
		}
	}
} else {
	fmt.Println("No privilege detected - safe to produce")
}

Endpoints

detect

Analyze document content for privilege indicators across multiple categories

Privilege categories

Category	Description
`attorney_client`	Communications between attorney and client for legal advice
`work_product`	Materials prepared in anticipation of litigation
`common_interest`	Shared communications under joint defense agreements
`litigation_hold`	Documents subject to preservation obligations

Recommendations

The API returns one of four recommendations based on analysis:

Recommendation	When returned	Suggested action
`withhold`	High confidence privilege detected	Do not produce; log for privilege log
`redact`	Partial privilege or mixed content	Redact privileged portions before production
`produce`	No privilege indicators found	Safe for production
`review`	Ambiguous or borderline cases	Requires attorney review

Vault integration

When you provide a document_id and vault_id, the API:

Retrieves the document content from your Vault
Analyzes it for privilege indicators
Stores the analysis results in the document’s privilege_analysis metadata field

This enables bulk privilege review workflows where results are persisted alongside documents.

# Analyze and store results in Vault metadata
casedev privilege:v1 detect \
  --document-id doc_abc123 \
  --vault-id vault_xyz789

# Retrieve the document with stored analysis
casedev vault:objects retrieve --id vault_xyz789 --object-id doc_abc123

// Analyze and store results in Vault metadata
const result = await client.privilege.detect({
  document_id: 'doc_abc123',
  vault_id: 'vault_xyz789',
});

// Results are now stored in the document metadata
// Retrieve later with vault.get()
const doc = await client.vault.get({ 
  vault_id: 'vault_xyz789',
  object_id: 'doc_abc123' 
});
console.log(doc.metadata.privilege_analysis);

# Analyze and store results in Vault metadata
result = client.privilege.v1.detect(
    document_id='doc_abc123',
    vault_id='vault_xyz789',
)

# Results are now stored in the document metadata
# Retrieve later with vault.get()
doc = client.vault.get(vault_id='vault_xyz789', object_id='doc_abc123')
print(doc.metadata.privilege_analysis)

// Analyze and store results in Vault metadata
result, _ := client.Privilege.V1.Detect(ctx, casedev.PrivilegeV1DetectParams{
    DocumentID: casedev.F("doc_abc123"),
    VaultID:    casedev.F("vault_xyz789"),
})

// Results are now stored in the document metadata
// Retrieve later with vault.get()
doc, _ := client.Vault.Get(ctx, "vault_xyz789", "doc_abc123")
fmt.Println(doc.Metadata.PrivilegeAnalysis)

Permissions

Input Type	Required Permissions
Raw text (`content`)	`chat`
Vault document (`document_id`)	`chat` + `vault`

Limits

Maximum document size: 200,000 characters
For larger documents, split into logical sections and analyze separately

Supported jurisdictions

Currently supports US-Federal privilege standards. Additional jurisdictions coming soon.

Vault

Store documents and run bulk privilege detection workflows

Legal Research

Verify citations in privileged documents before logging

OCR

Extract text from scanned documents before privilege analysis

Get Started

Platform

Resources

Privilege Detection Overview

The Problem

The Solution

Quick start

Endpoints

detect

Privilege categories

Recommendations

Vault integration

Permissions

Limits

Supported jurisdictions

Vault

Legal Research

OCR

​The Problem

​The Solution

​Quick start

​Endpoints

detect

​Privilege categories

​Recommendations

​Vault integration

​Permissions

​Limits

​Supported jurisdictions

​Related services

Vault

Legal Research

OCR

The Problem

The Solution

Quick start

Endpoints

Privilege categories

Recommendations

Vault integration

Permissions

Limits

Supported jurisdictions

Related services