Jointli Book a scoping callScoping call

AI acceptable use policy template

An AI acceptable use policy is a short document that tells your team which AI tools they can use for work, what they can put into them, and who checks what comes out. If anyone at your firm uses ChatGPT, Copilot, Gemini or Claude for work, this is the page that turns that habit into something you can explain to a client, an insurer or an examiner.

Why it matters now

Picture a Tuesday. Someone on your team is behind on a client report, pastes the client's figures into a free AI chatbot to tidy the wording, and sends it out by noon. The report reads well. Nobody checked the numbers the chatbot rewrote, and nobody knows where the client's figures went.

A one-page policy prevents both problems. It names the tools that are approved for client work, it keeps sensitive information out of the rest, and it makes sure a person checks every AI-assisted piece before it leaves the building. When a client's vendor questionnaire or your cyber insurance renewal asks how you handle AI, this is the document you point to.

What the template covers

The sections line up with the risks NIST names in its Generative AI Profile (NIST AI 600-1), which builds on the NIST AI Risk Management Framework:

Data privacy: what never goes into an AI tool (Section 4), keeping each client separate (Section 5), and how accounts are set up (Section 9).

Confabulation, NIST's term for confident answers that are wrong: a named person checks every output and notes the check (Section 6).

Information security: approved accounts, settings, logs and new-tool reviews (Sections 9 and 10).

Intellectual property: nothing goes in that you don't have the right to share (Section 4).

Over-reliance on AI: each tool's list says what it must not be relied on for, and people make the decisions (Sections 3, 6 and 7).

Third-party tools: four terms are checked in writing before a tool touches sensitive information (Section 10).

The template

Email us and we'll send you the free template as a Word file you can edit. It has twelve short sections, with brackets where your firm's details go. Most firms can fill it in during one meeting, which saves you from writing one in a rush the week a client's questionnaire asks for it.

What's inside

  1. Purpose
  2. Who it covers
  3. Approved tools
  4. What never goes into an AI tool
  5. Keep each client separate
  6. A person checks every output
  7. Decisions about people
  8. Being open about AI
  9. Accounts, settings and logs
  10. Adding or changing a tool
  11. When something goes wrong
  12. Training, signatures and review

This template is a starting point, not legal advice. Have your counsel review it against your industry's rules before you adopt it.

Put the policy in place this week

List the tools your team already uses. Ask everyone, including the tools they use on their phones. That list becomes Section 3.

Fill in the brackets. Name one approver and one contact for problems. In a small firm that can be the same person.

Set the accounts up. Move work onto company accounts, switch on the privacy settings each tool offers, and check the four vendor terms in Section 10.

Walk the team through it. Fifteen minutes in a staff meeting covers it, with time for questions.

Collect signatures and set the review date. Put the next review on the calendar now.

A team lead walking two colleagues through a printed AI policy

For law firms

The template works for a law firm. The American Bar Association's Formal Opinion 512 on generative AI (July 2024), which explains the ABA Model Rules, points to four places to tighten it. Check your own state's rules as well.

Client consent (Section 4). Before client information goes into an AI tool that learns from what it's given, the client's informed consent is needed. The opinion says general boilerplate in an engagement letter is not enough.

Read the tool's terms (Section 10). Each lawyer using a tool should read its terms of use and privacy policy, or rely on a colleague or outside expert who has. Our vendor due diligence questionnaire covers the rest of the check.

Check before it's filed (Section 6). Review every AI-assisted output for accuracy, above all before anything goes to a court. Citations to cases that don't exist are among the problems that have already come up.

Set the policy and train (Section 12). Lawyers who manage a firm must set clear rules for how AI may be used, and make sure lawyers and staff are trained. This template is that policy.

Questions firms ask

If even one person uses AI for client work, yes. It is the simplest written answer when a client or insurer asks how you handle AI, and it takes one meeting to put in place.

No. The policy is the set of rules your team follows. AI governance is how your firm decides which AI to use, approves it, watches how it performs, and keeps a record of those decisions over time. The policy is one piece of that.

Your policy decides. This template says no for company work by default, because personal and free accounts sit outside your firm's control: you cannot set their privacy options or see what was entered.

Every six months, and whenever you add a tool or a tool changes. AI products change their features and terms often, so a scheduled review keeps the policy matched to what your team actually uses.

When a policy isn't enough

A policy says what your team should do. An AI risk assessment checks what is really happening: which tools are in use, where your information goes, and what your clients, contracts and regulators expect of you. Jointli runs it live with your team, with the owners on every call, and sets the fee before we start.

Ask us about an AI risk assessment →

Built on the NIST AI Risk Management Framework (AI RMF 1.0, January 2023) and its Generative AI Profile (NIST AI 600-1, July 2024).