AI tools and client data

Staff AI policy builder

Answer three things — what kind of practice this is, which AI tools are approved and on what plan, and which information must stay out of an unapproved tool — and this page assembles a one-page staff AI policy you can copy. The clauses are the ones from our staff AI policy article, unchanged. It runs in your browser: nothing is sent to us and nothing is saved.

Three answers and it writes itself

What kind of practice is this?

Which AI tools are approved, and on what plan?

These become the approved-tool register inside the policy. Leave a tool blank to leave it out.

  • ChatGPTGeneral assistant
  • Microsoft CopilotGeneral assistant
  • Google GeminiGeneral assistant
  • ClaudeGeneral assistant
  • PerplexitySearch assistant
  • DeepSeekGeneral assistant
  • Zoom AI CompanionMeeting notetaker
  • Otter.aiMeeting notetaker
  • Fireflies.aiMeeting notetaker
  • Read AIMeeting notetaker
  • GrammarlyWriting aid
  • Notion AIWorkspace assistant
  • GitHub CopilotCoding assistant
  • CursorCoding assistant
  • Claude CodeCommand-line agent
Which information must stay out of an unapproved tool?

Your policy

Nothing is sent anywhere and nothing is saved. Copy it when it says what you mean.

A client confidence pasted into a personal account has left the firm's control. The decision is whether a personal account may touch a client matter at all. what we build for law firms.

This is an operational starting point, not legal advice and not proof of compliance: approve it with the people responsible for privacy, professional obligations and IT, and adapt it to your practice. The clauses come from our staff AI policy article, which carries the sources. Writing the list is easier once you know what people already use: a free 30-day AI security assessment takes that inventory off the computers.

A policy is the second step

The first is knowing what people already use, because an approved-tool register written from memory approves the wrong list. The AI tool risk checker works through the tools you know about, and a free 30-day AI security assessment finds the ones you do not. A sample set of findings shows what comes back.