Skip to main content
Security and ethics

AI and client confidentiality: what lawyers must protect

To protect client confidentiality when using AI, decide what information may be entered, use only tools whose data terms you have checked, remove identifiers where you can, restrict access to the matter, and have a lawyer review all output. If a tool may retain or learn from inputs, treat it as a disclosure risk.

By Raghav Handa · Updated 17 October 2026 · 3 min read

Quick decision table

InputUnapproved consumer toolApproved tool with checked terms
Public law or generic questionUsually fineFine
Anonymised hypotheticalUsually acceptable with careFine
Client facts, names, documentsAvoidPer policy; consent may be needed
Privileged adviceAvoidRestricted; senior approval

What to check in the terms

  • Is customer content used to train models?
  • How long are prompts and outputs retained?
  • Who are the subprocessors and where do they operate?
  • Can the provider's staff review content?
  • What happens on termination?
  • Can admins control or disable features?

Terms differ by plan and change over time. Record the date you checked.

Safe prompting habits

  • Replace names with roles such as Client A.
  • Strip account numbers, addresses and dates of birth.
  • Give the minimum text needed.
  • Do not upload entire files unless the tool is approved.
  • Keep a note of what you entered, for your records.

Inside Notion, the same logic applies to Notion AI: it acts on content a user can reach, so keep privileged pages restricted. LawyerOS Pro's AI Agents page offers prompt patterns but does not change Notion's data handling.

General information only, not legal advice. Check the rules of your own bar or regulator and your clients' terms.

Frequently asked questions

Can I paste client emails into ChatGPT?
Only if your firm has approved that tool and plan and you have met any consent requirements. Many firms prohibit consumer tools for client data.
Is anonymising enough?
It reduces risk but facts can still identify a client. Use judgment.
What about AI notetakers?
Recording and transcribing calls raises confidentiality and consent issues. Check privilege, local recording laws and vendor terms.
Does an enterprise plan solve it?
It may offer better terms but you still need to read them and configure access.

About the author

Raghav Handa

Raghav Handa trained as a lawyer, then spent years building legal technology, from enterprise legal products to custom systems for law firms.

Want it built for you?

LawyerOS is a ready-made Notion workspace for matters, deadlines, time and invoices. Starter $19, Pro with AI $39, one-time.

See LawyerOS