Why write it down
Everything in this course comes down to a small number of decisions: which tools, for which material, for which tasks, with which checks. You can make those decisions afresh every time you open a chat window, under time pressure, at the end of a long day. Or you can make them once, calmly, write them down, and follow them.
The written version is better for three reasons. It is more consistent, because the same question gets the same answer. It is easier to explain, if a client, a supervisor or a regulator ever asks how you use AI. And it is easier to share, so a team works to one standard instead of five private ones.
This lesson helps you draft a short personal or team policy. It sits underneath your firm's policy, never above it. If your firm already has one, your version is a working summary of how you apply it. If it does not, your draft is a sensible thing to take to whoever is responsible for risk and ask them to review.
A policy you draft here is a working document, not legal advice and not a substitute for your firm's policy or your regulator's guidance. Check it against your own regulator, bar or law society guidance, the rules of the courts you appear in, and your firm's policy, and have it reviewed by whoever is responsible for risk in your organisation.
What a good policy covers
Keep it short enough to read in two minutes. A policy nobody reads protects nobody. Write it in plain language, as instructions to yourself and your colleagues rather than as a legal document. It should answer seven questions.
| Question | What your answer should say |
|---|---|
| Which tools? | The tools approved for client material, and what may go into anything else |
| Which material? | What may be shared, what may not, and when to redact |
| Which tasks? | Where AI is used and where it is not, such as advice and authorities |
| Which checks? | How output is verified before anyone relies on it |
| Who is responsible? | Who signs off work that used AI, and who owns the policy |
| What to disclose? | How you find and follow court, regulator and client requirements |
| What if it goes wrong? | Who to tell, how quickly, and what information to give |
A template to start from
Copy this, fill in the brackets from your firm's policy and your own practice, and delete anything that does not fit.
LEGAL AI POLICY: [name or team]
Version [number], [date]. Owner: [name]. Review by: [date].
This sits under [firm policy name]. Where they differ, the firm policy wins.
-
TOOLS
Approved for client or matter material: [tools, as approved by the firm].
Other tools: no client, matter or personal data. General use only.
-
MATERIAL
Never paste into any tool without specific approval: [list, e.g.
privileged advice, settlement figures, health or other sensitive personal data].
Before pasting into an approved tool: share only what the task needs,
and redact names and identifying facts where the task allows.
-
TASKS
We use AI for: first reads and summaries, routine first drafts from
our own precedents, chronologies with source references,
plain-English explanations, checklists, research search terms.
We do not use AI for: legal conclusions, advice, authorities,
or anything that has not been checked by a qualified person.
-
CHECKS
Every authority, quotation and statement of law is read in the
primary source before it is filed, sent or relied on.
Every summary point we rely on is checked against the source.
Every draft is reviewed clause by clause, including defined terms.
-
RESPONSIBILITY
The person whose name is on the work is responsible for all of it,
however it was produced. Supervisors check AI-assisted work from
juniors as they would any other work.
-
DISCLOSURE
Before filing, check the current rules and guidance of the court
or tribunal on AI use. Check client guidelines and engagement terms
before using AI on a matter. Where unsure, ask [name or team].
-
MISTAKES
If material goes into the wrong tool, or AI-generated error reaches
a client, court or other party, tell [name or team] promptly
with what happened, where and when. Do not try to fix it quietly.
Checkpoint
A short written policy on tools, material, tasks, checks, responsibility, disclosure and mistakes makes your AI use consistent and explainable, and it always sits under your firm's policy.
Making it stick
Keep it visible. Pin it where you work. The moment it matters is the moment you are about to paste something, not the moment you are reading a policy folder.
Date it and review it. Tools, terms and professional guidance change. Put a review date on the policy and keep it. When your firm or regulator issues new guidance, update your version the same week.
Use it in supervision. When a junior colleague brings you AI-assisted work, walk through the checks section together. It teaches the habit and shows that the standard applies to everyone, including you.
Record the important uses. For significant pieces of work, a short file note of which tool was used, for what and how the output was checked may be useful later. Your firm's policy may say whether and how to record this.
Below is a draft personal AI policy for legal work.
Do not tell me what any law or professional rule requires.
Instead, act as a sceptical colleague and list:
- Situations the policy does not clearly cover
- Places where two sections could conflict
- Wording that someone could reasonably read two ways
- Questions I should put to my firm's risk team before finalising it
Policy:
[paste your draft, with no client information]
Where to go next
If you want to go deeper, these courses build on what you have covered here:
- AI Safety, Privacy and Verification for a fuller treatment of checking output and handling sensitive information.
- Safe AI Adoption for Teams if you are responsible for how a team or department uses AI.
- AI for Research and Study for more on search terms, source checking and invented references.
- Free AI courses with a certificate to keep building your skills.
Checkpoint
Keep the policy visible, dated and reviewed, and use it in supervision so the standard is shared across the team.
๐ Quiz
Question 1 of 4How should your personal or team AI policy relate to your firm's policy?