Appearance
Legal
GhostAI provides specialized protection for law firms and legal departments handling privileged and confidential information.
Overview
Legal organizations face unique challenges:
- Attorney-Client Privilege: Must be strictly maintained
- Work Product Doctrine: Protect legal strategies
- Client Confidentiality: Professional responsibility
- Matter Isolation: Prevent cross-matter contamination
Legal Shield Add-On
The Legal Shield add-on provides enhanced privilege protection, matter isolation, legal document intelligence, and profession-specific compliance support. Available on Personal and above at $12/user/mo.
Quick Start for Legal
1. Select Legal Vertical
In Settings → Industry Vertical, select Legal.
This enables:
- Enhanced client name detection
- Matter/case number protection
- Privilege-aware processing
- Legal document recognition
2. Enable Legal Detection
| Data Type | Examples |
|---|---|
| Client Names | Individuals and entities |
| Case Numbers | Court docket numbers |
| Matter IDs | Internal matter numbers |
| Bates Numbers | Discovery references |
| Confidential Codes | Settlement amounts |
3. Configure Matter Isolation
Enable matter-based data isolation:
- Separate token vaults per matter
- No cross-matter data leakage
- Matter-specific audit logs
Key Features
Privilege Protection
GhostAI helps maintain privilege by:
- Tokenizing client identifiers
- Protecting case strategies
- Isolating matter communications
- Logging access for privilege logs
Document Intelligence
- PDF Generation for contract summaries, NDA reviews, legal briefs, and due diligence checklists — with rich tables (clause comparisons, term matrices) and blockquotes (legal citations, statutory references)
Recognize and protect:
- Pleadings and motions
- Discovery documents
- Contracts and agreements
- Correspondence
Public Figure Reconciliation
Names of judges, legislators, and public officials referenced in case law and legal research (e.g., "Justice Sotomayor", "Senator Warren") are recognized as public figures and excluded from PII detection, reducing false positives in legal drafting and analysis. Co-occurrence and multi-detector safety filters ensure that client names are never mistakenly reconciled, even when they share a name with a public figure.
Public Knowledge Reconciliation
Legal teams can reconcile public ORGANIZATION, LOCATION, and EVENT entities in legal research and drafting. ORGANIZATION uses a curated list (and optional tiers), not a registry of public companies. Client/matter-private context remains protected by co-occurrence and contextual safety filters.
Conflict Checking Support
AI assistance without exposing:
- Potential client names
- Adverse party information
- Related entity connections
Use Cases
Contract Review
Scenario: Associate reviewing commercial agreement.
Input:
Review this agreement between:
Client: Acme Corp
Counterparty: XYZ Industries
Key terms: $5M deal, 3-year term
Unusual clause on page 15 regarding IPProtected: Party names tokenized; AI reviews terms.
Brief Writing
Scenario: Drafting motion for client.
Input:
Draft summary judgment motion for:
Client: John Smith, Case No. 24-CV-1234
Key facts: Breach of contract, damages $500K
Defendant: ABC Company failed to deliverProtected: All party names and case numbers secured.
Due Diligence
Scenario: M&A transaction review.
Input:
Summarize due diligence findings:
Target: TechCo Inc
Acquirer: BigCorp
Key issues: IP ownership, pending litigation
Deal value: $50MResult: AI summarizes with protected party information.
Compliance Features
Professional Responsibility
GhostAI supports compliance with:
- ABA Model Rules
- State bar requirements
- Client confidentiality duties
Audit Trail
Complete logging for:
- Ethics investigations
- Malpractice defense
- Privilege assertions
Best Practices
- Enable matter isolation: Prevent cross-matter exposure
- Use strict thresholds: Legal data requires high sensitivity
- Review audit logs: Monitor access patterns
- Train associates: Ensure understanding of protections
- Document policies: Maintain compliance evidence
Memory & Attorney-Client Privilege
AI Memory in legal deployments includes safeguards to protect attorney-client privilege and work product:
- Matter-Scoped Memories: Memories can be scoped to specific matters using the conversation and subject memory scopes, preventing cross-matter contamination. Each matter's memories are isolated in the same way as matter token vaults.
- Privilege Protection: All memories pass through legal-specific PII detection before storage, tokenizing client names, case numbers, matter IDs, and Bates numbers. Privileged content in memories receives the same protection as chat messages.
- No Cross-Matter Leakage: Organization and group memory scopes enforce matter isolation boundaries. A memory from Matter A is never surfaced in Matter B, even within the same firm.
- Audit Trail for Privilege Logs: Every memory access is logged with sufficient detail (who, when, what matter) to support privilege log preparation and ethics investigations.
- Retention & Destruction: Memory retention policies can be configured per-matter to support litigation hold requirements. When a matter concludes, associated memories can be bulk-purged with documented chain of custody.
- Work Product Doctrine: AI-generated memories derived from legal analysis are flagged as potential work product, helping firms maintain doctrine protections.
- Memory Interleave (privilege): Memory Interleave can be disabled for privileged conversations. Matter-scoped recall ensures memories don't cross case boundaries.
Company Intelligence Profile for Legal
The Company Intelligence Profile strengthens legal detection:
- Matter names and client codenames: Register active matter names, client codenames, and case identifiers as sensitive terms
- Opposing counsel: Add opposing party names and counsel identifiers to the sensitive registry
- Internal project names: Include firm initiative names, practice group identifiers, and internal billing codes
- Safe terms: Register firm name variations, public case citations, and standard legal terminology that generate false positives
- Auto-Research: One-click profile population identifies your firm's practice areas and suggests relevant terms
Fine-tuning detection for Legal
To improve detection for legal-specific entities (e.g. matter IDs, client identifiers, case numbers), you can fine-tune the PII model on your labeled data. See PII Fine-Tuning and the True ML Fine-Tuning whitepaper for data format and per-vertical examples.