Documentation & Communication: Make AI Reviewable

AI Ethics for Legal Professionals

Documentation & Communication: Make AI Reviewable

If you use AI without documentation, you create blind spots. If you document poorly, you create noise. The goal is a lightweight audit trail.

What to log (minimum viable audit trail)

Tool (name, version, deployment type)
Date/time and matter reference
Purpose (what you asked the tool to do)
Inputs (sanitized summary, not raw secrets)
Output (saved copy or link)
Verification steps (what you checked)
Reviewer (who signed off)

A three‑phase checklist

{{UPLOAD_ASSET:secure_ai_use_triangle.png}}

Triangle diagram showing secure AI use before during after — *A simple three-phase checklist: before you use AI, while you use it, and after you use it.*

Communication template: message to the supervising attorney

Use this short format when requesting review:

AI‑assisted draft ready for review
Task: [e.g., summarize deposition for internal memo]
Tool: [Approved Tool]
Inputs: sanitized; no privileged facts pasted into public systems
What I verified: [facts X/Y; citations A/B; dates; names]
Open issues / questions: [list]
Risk tier: [low/moderate/high] (my estimate)

Activity: create your “done definition”

Pick one recurring task (e.g., summarizing discovery responses). Define what “done” means:

What must be verified?
What must never be included in prompts?
Who must approve before it goes out?