Dr. John Smith at Acme Inc. asked us to review the $450,000 settlement proposal. Send comments to [email protected]. Project Phoenix must remain confidential.
Use case 05 · Legal documents
Review sensitive legal details
before they reach AI.
Create a reviewable working copy that replaces direct identifiers and confidential terms before legal text is used with ChatGPT, Claude, Gemini, or another approved AI tool.
The practical problem
A contract, brief, or matter document could benefit from AI, but the source contains party names, contact details, transaction values, and matter-specific information.
The AI receives contextual tokens instead of the original values, while the person handling the matter keeps control of the local mapping.
A protected handoff
The useful context remains.
Selected details are replaced.
Dr. [[PERSON_1]] at [[COMPANY_1]] asked us to review the [[AMOUNT_1]] settlement proposal. Send comments to [[EMAIL_1]]. [[CUSTOM_1]] must remain confidential.
Replace locally
What gets protected
- People and organizations
- Email addresses and phone numbers
- Financial values
- Matter names and custom terms
Keep useful
What the AI can still understand
- Who did what
- The type of legal document
- The instruction and requested task
- Repeated relationships between entities
Example task
A narrow prompt produces a more controlled handoff.
“Summarize the proposal, identify open issues, and draft a neutral list of questions for counsel to review.”
Send only the protected working copy, and only through an AI provider and account approved for this work.
Human review remains essential
Protection is a checkpoint, not a permission slip.
- 01
Add matter names or client aliases the detector cannot infer.
- 02
Look for distinctive facts that could still identify the matter.
- 03
Confirm the AI provider and account are approved by the firm.
Ready when you are
Make the protected copy before the AI conversation starts.
Open the matching synthetic example, review every replacement, and see the complete local workflow.