Prompting With Role Guardrails (Templates You Can Reuse)
You can reduce UPL risk by using prompts that keep the AI in a support role. Your prompts should ask for organization, options, and citations—not legal advice.
Role‑safe prompt wrapper
Use this wrapper at the top of prompts for staff‑level tasks:
ROLE: You are an AI drafting assistant. You do not provide legal advice. You only help organize information and draft text for attorney review.
CONSTRAINTS: If asked for legal advice, respond with: “This requires attorney review.” Always cite sources when making legal assertions. Flag uncertainty.
OUTPUT: Provide a structured draft plus a verification checklist.
Example: deposition summary (safe)
Task: Summarize the attached deposition transcript for internal use.
Constraints: Do not make legal conclusions. Extract key facts, dates, and contradictions. Use neutral language. Provide a list of questions for attorney follow‑up.
Example: client email draft (high‑risk; requires sign‑off)
Task: Draft an email that explains next steps in plain English based on this attorney‑approved outline.
Constraints: Do not add new advice. Do not interpret the law beyond the outline. Use a professional tone. Mark any sentence that should be confirmed by the attorney.
Verification checklist (copy/paste)
- Did the output add new advice beyond what was provided?
- Did it cite law/cases? If yes, verify each citation.
- Does it contain any confidential client facts that should not be shared?
- Is attorney review required before release?
Activity: turn one of your prompts into a guarded prompt
Take a real prompt you used recently. Add:
- Role statement (no advice)
- Constraints (cite sources, flag uncertainty)
- Verification checklist request