Sample AI Policy for Law Firms: What to Include

Start with Lawcover's Sample AI Policy or the QLS template, check it against the December 2024 joint regulator statement and the practice notes of the courts the firm appears in, then have the principal approve it. I would put the approved tools, verification steps and records ahead of polishing the wording. This article is general information, not legal advice.

Where Can an Australian Law Firm Get a Sample AI Policy?

I checked these two starting points on 6 November 2026. I could not read the text of either template, so I cannot assess their clauses.

Lawcover's page shows no fee and the file downloaded without a login, but I could not confirm that either template is free. The real cost is principal time to assess the tools, adapt the document and train staff. I would choose the sample that needs the least change for the firm's courts and state. The QLS listing's April 2023 date predates the December 2024 joint statement, so that template needs checking against later rules.

Choose the permitted work alongside the policy. Truespeak's guide to what law firms should automate first helps frame that decision.

What Clauses Should a Law Firm AI Policy Include?

The following is Truespeak's plain-English starting outline, not wording from either template or a legal precedent. The principal should adapt it to the firm's state and courts and take their own professional advice before adoption.

The basis is the December 2024 joint regulator statement, QLS companion guide and court guidance below.

1. Purpose and scope. Cover staff, principals and contractors, including AI built into software already used by the firm.
2. Approved tools. Maintain a register with a named owner, approved settings and an approval pathway. Require approval before accepting tool terms.
3. Permitted work. Specify who may use each tool and for which tasks. Start with lower-risk work that a qualified person can readily verify.
4. Prohibited uses. Prohibit confidential, sensitive or privileged client information in public AI tools, material under suppression or non-publication orders, and replacing a lawyer's advice or judgement.
5. Tool assessment. Check confidentiality, security, privacy and contractual terms before approval. Define what information each approved tool may receive.
6. Verification. Require a qualified person to check each output against original sources before reliance. Another AI tool does not count as verification.
7. Court work. Require the responsible lawyer to check the current court rules before filing, including evidence, expert reports and disclosure.
8. Supervision and training. Train users before access and assign active supervision of junior and support staff.
9. Clients and billing. Define when clients are told about AI use and keep time entries accurate to the work performed.
10. Records and review. Record the tool, prompts, outputs and checks under the firm's retention and access rules. Set an error-reporting process and review date.

The error-reporting process and review date are my practical suggestions. For client information, read the OAIC guidance alongside Truespeak's guide to putting client data into AI tools. OAIC recommends against entering personal, particularly sensitive, information into publicly available generative AI tools.

What Has Happened to Australian Lawyers Using AI, and Has Anyone Been Struck Off?

In Dayal, the Victorian regulator varied the practising certificate on 19 August 2025 after inaccurate authorities were tendered. The restrictions prevented principal practice and handling trust money, while permitting employee practice with two years of supervision and quarterly reporting.

Mills Oakley reports that the AI tool was incorporated into practice management software. That is why I would expressly cover built-in AI.

In Murray v Victoria [2025] FCA 731, Mills Oakley reports insufficient care in research and inadequate supervision. The firm was personally ordered to pay indemnity costs.

I could not find an Australian lawyer struck off for AI misuse as at 6 November 2026. In England and Wales, a foreign lawyer was struck off the register of foreign lawyers in a matter involving AI-generated false citations. Aldwych Legal reports other grounds would have justified removal without the AI issue.

Do Lawyers Have to Tell Courts and Clients When They Use AI?

The answer depends on the court and the work. Start with the Law Society of NSW court protocols index, dated 14 May 2026, then check the court's current text for later changes.

For NSW Supreme Court work, check SC Gen 23. Barry Nilsson's February 2025 explanation describes restrictions on generating affidavit and witness statement content, and prior-leave requirements for expert reports.

For Federal Court work, Hall & Wilcox explains that GPN-AI requires disclosure in some circumstances. So a firm's policy should tie disclosure to the circumstances the practice note describes.

For clients, QLS Guidance Statement 37 says clients should be advised where feasible and time-based billing must remain accurate. The joint statement also calls for transparency, recording and client disclosure.

Who Should Approve AI Tools and Make the Policy Work Daily?

I would have the principal approve the policy and name a senior person to maintain the tool register. The QLS companion guide suggests a small firm consider tasking a senior person to maintain an overview of the AI tools in use.

1. List existing tools and built-in AI features.
2. Assign permitted tasks, reviewers and data limits.
3. Train staff on approval and error reporting.
4. Check records and revisit permissions when tools or rules change.

Truespeak's Australian law firm AI tools guide can inform the inventory; product selection still needs the firm's assessment.

What Are the Red Flags in a Sample AI Policy?

I would question overseas rules copied unchanged, court references that do not match the practice, no named approver, silence on built-in AI and no verification record. An older date calls for checking against later rules. A blanket ban staff cannot apply needs a workable approval pathway.

Where Does Truespeak Fit After Policy Approval?

Truespeak is the best fit for an Australian law firm that has set its AI rules and wants workflows designed, built and run with a person approving anything sensitive and the outcome recorded. I spent 20+ years in commercial roles, including at Optus, Samsung and Nikon, and I have built 25 working productivity systems.

I do not give legal advice or write or approve the firm's legal policy. On Truespeak's AI for law firms page, I set out how those approved rules can become workflow steps: prepare an action, hold it for approval and record the outcome. Truespeak's human approval guide explains that approach.