Free template · Law firm AI policy
Law firm AI policy template, in plain English.
What should a law firm AI policy cover?
A law firm AI policy should say which tools people may use, what client information can go into them, that a lawyer checks every output before it is relied on, what AI is never used for, how it is billed, and who is responsible. The ten sections below cover all of it. Copy them, fill in the brackets, and make them yours.
A starting point, not legal advice. Check your state bar's guidance and your courts' rules.
The template
1. Purpose and scope
This policy covers every use of artificial intelligence tools by anyone working for [Firm Name], including attorneys, staff, contractors and interns, on any device, for any firm or client work.
Why it is here: A policy that only covers some people or some devices is the gap someone falls through.
2. Approved tools
Only the tools on the firm's approved list may be used for firm work: [list tools]. The approved list is kept by [name or role]. To add a tool, ask [name or role] first. Personal accounts on public AI tools may not be used for client work.
Why it is here: Most AI mistakes at firms start with a well-meaning person pasting something into a free tool nobody vetted.
3. Client confidentiality
Client names, facts, documents and any other information relating to a representation may only be entered into an approved tool that the firm has confirmed does not use our data to train its models and protects it by contract. When in doubt, remove identifying details or ask first. Before client information goes into a tool that learns from what it is given, the client's informed consent is required, and a general line in the engagement letter is not enough.
Why it is here: Model Rule 1.6. ABA Formal Opinion 512 puts confidentiality at the center of any AI use.
4. Check everything before it is used
Every AI output is a draft. Before anything produced with AI is relied on, sent to a client or filed, a lawyer must verify it against the source. Every case citation must be pulled and read. No citation may be filed that a lawyer has not personally confirmed exists and says what we say it says.
Why it is here: Model Rule 1.1 (competence). In Mata v. Avianca, a federal court in New York sanctioned lawyers $5,000 on June 22, 2023 for filing a brief with cases that did not exist.
5. What AI is never used for
AI is not used to give legal advice to a client, make settlement decisions, handle trust money, sign or file anything with a court without a lawyer's review, or communicate with a client in the firm's name without a person approving the message. [Add the firm's own items.]
Why it is here: Writing down what stays human is the part most policies skip, and it is the part clients and courts care about most.
6. Talking to clients about AI
When AI use is material to a client's matter, or the client asks, we tell them how we use it. Engagement letters will state [the firm's chosen language].
Why it is here: Model Rule 1.4 (communication), as discussed in ABA Formal Opinion 512.
7. Billing
We bill for the time a lawyer spends using an AI tool on a matter and reviewing what it produced. We do not bill clients for time spent learning a tool. General AI subscriptions are firm overhead, not a client expense. Any other AI cost passed to a client is agreed with the client in advance.
Why it is here: Model Rule 1.5 (fees). ABA Formal Opinion 512 uses the learning-time example directly and treats general tool costs as overhead.
8. Supervision and training
Lawyers with managerial or supervisory authority make sure attorneys, staff and outside services follow this policy. Anyone who uses an approved tool is trained on it first, including what it gets wrong.
Why it is here: Model Rules 5.1 and 5.3 make supervising lawyers responsible for the lawyers, staff and outside services they rely on.
9. Courts
Before filing in any court, check that court's standing orders and local rules for AI disclosure or certification requirements, and follow them.
Why it is here: A growing number of judges require a statement about AI use in filings. The rule is set court by court.
10. Mistakes and review
If client information goes into an unapproved tool, or an AI error reaches a client or a court, tell [name or role] the same day. This policy is reviewed every [six months] and whenever the firm adds a tool.
Why it is here: A mistake reported the same day is a fix. A mistake found later is a problem.
Where these rules come from
Most of the template follows ABA Formal Opinion 512, issued July 29, 2024. It does not invent new duties. It applies the ones you already have to AI: competence, confidentiality, communication and fees.
Section 4 exists because of cases like Mata v. Avianca, where a federal court in New York sanctioned lawyers $5,000 on June 22, 2023 for filing a brief citing cases that did not exist. The tool made them up. Nobody checked.
From a policy to a plan
A policy says what not to do. The harder question is where AI should actually help at your firm, and that is different for a two-lawyer injury practice than for a twenty-lawyer business firm.
That is what our free AI Map is for: a 30-minute call where you walk us through how your firm runs, and you leave with one page showing where AI can do the heavy lifting, where it should never touch, and what to do first. The "never touch" column drops straight into section 5 of this template.
Common questions
Does my law firm need an AI policy?
If anyone at your firm uses AI for work, yes, even if the policy is one page. Your people are almost certainly using it already. A policy turns that from a risk nobody is managing into a rule everybody knows.
Is this template legal advice?
No. It is a starting point written in plain English. Your state bar may have its own guidance, and some courts have their own rules, so read those and adapt the template to your firm.
What does ABA Formal Opinion 512 require?
Issued July 29, 2024, it is the ABA's first ethics opinion on generative AI. It says the existing rules apply, including competence with the tools you use, protecting client information, communicating with clients, reasonable fees, supervising others, and candor to courts.
Can staff use ChatGPT at a law firm?
Only under clear rules. The main risks are client information going into a tool that is not protected by contract, and output being trusted without checking. Sections 2, 3 and 4 of the template cover exactly that.
Ready to put AI to work, safely?
Tell us how your firm runs. We will show you where AI fits, where it does not, and what it would take to build.
Book a $250 ConsultThe consult: one hour with the team, not a salesperson. You leave with a real number. The $250 comes off your build if you move forward. You pick your time right after checkout.
Not ready to pay yet? Get a free 30-minute AI Map for your firm.