Avoiding Unauthorized Practice of Law (UPL) in the Age of AI
AI can blur boundaries because it produces “legal-sounding” text. The key risk is when a non-attorney uses AI output to advise, interpret, or recommend in ways that cross into legal judgment.
Why AI increases UPL risk
- AI produces persuasive legal language—even when it is wrong or incomplete.
- Staff may be tempted to answer client questions quickly using AI summaries.
- AI can generate strategy (“you should file X”) that feels actionable.
What usually stays safe for staff (with supervision)
- Summarizing documents or transcripts without adding conclusions
- Extracting dates, parties, and defined terms
- Drafting checklists and procedural to-do lists for attorney review
- Formatting drafts and organizing exhibits
What usually crosses the line
- Telling a client what they “should” do or which legal option is “best”
- Interpreting law or applying law to facts as a final conclusion
- Negotiation positions, settlement advice, or threats
- Privilege determinations and legal conclusions without attorney review
Role boundary guardrails (simple rules)
- No direct client advice. Route client questions to an attorney.
- Label as draft. AI output is “draft for attorney review.”
- Use templates. Prompts should include role limits and “do not provide legal advice” constraints.
- Escalate ambiguity. If the task requires judgment, stop and escalate.
Practice tip: A safe staff output reads like “Here is what the document says,” not “Here is what you should do.”