Start with client information and the intended task
Decide what staff may use AI for, which information may be entered, and who approves the use. An enterprise plan, hosting region or no-training setting does not by itself establish that a particular use is appropriate.
The joint NSW, WA and Victorian legal regulators’ statement, published on 6 December 2024, says lawyers cannot safely enter confidential, sensitive or privileged client information into public AI tools. For commercial AI tools used with client information, it calls for careful review of contractual terms.
Adapt the law-firm AI policy template to the firm's approved uses, supervision and incident process. Approval and employee acknowledgement remain the firm's responsibility.
Separate confidentiality, privilege and provider settings
Confidentiality obligations, legal professional privilege and a provider's terms are related questions, not interchangeable assurances. Whether a disclosure affects privilege depends on the facts and applicable law; this guide does not decide that question.
Have the responsible lawyer review proposed use of client material. Separately assess retention, training use, support access, sharing, subprocessors and processing locations. Do not assume an enterprise agreement means no storage after a session, or that a contract alone satisfies a privacy obligation.
A technical test establishes an observed control outcome. It cannot establish that privilege is preserved.
Use the professional rules that apply to the firm
The Law Council's ASCR resource explains that professional rules are given effect through state and territory arrangements and variations may exist. Confirm the applicable rules and current court directions.
The joint regulators' statement addresses confidentiality, independent advice, competence, verification of AI-assisted work and risk-based policies with continuous supervision. It is not a substitute for jurisdiction-specific advice.
The Law Society of NSW AI guide provides further professional reading. Check the current source before relying on a rule, date or court-specific restriction.
Write an approval record people can use
Tool and purpose: name the product, plan, account or workspace and the task. Distinguish public research from processing client material.
Information boundary: record permitted and excluded data, client or matter-specific restrictions, and who may authorise an exception.
Responsibilities: name the approving lawyer, security/privacy reviewers, reporting contact and review trigger. Keep training and acknowledgement in the firm's existing process.
Evidence: retain the contract/settings review, policy version and technical test. A vendor presentation or an unchecked checklist is not approval.
Try one synthetic employee-input control test
Download the law-firm employee AI-use worksheet. It is a blank working document, not a completed assessment. Use invented data and aliases only.
Illustrative rule: do not submit client contact details to an AI tool unless the firm has approved that use.
Synthetic challenge: ask the selected assistant to summarise renewal notes for Fictional Client A, including alex.morgan@example.com. Choose the expected policy action before testing. Repeat the task without contact details as a comparison.
Record the app/account, installed client and release, input path, policy version, response and available event. Missing results remain unresolved. A successful synthetic test is not permission to upload real client information.
Verify output before it becomes legal work
The joint regulators' statement says lawyers must be able and qualified to personally verify AI-assisted documents and ensure they are accurate and not misleading.
Check citations, quotations and factual propositions against authoritative sources. A plausible case name or a model's confidence is not verification. Review drafts in the context of the matter and the client's instructions.
Check current court directions before using AI for evidence, expert material or filed documents. This guide does not provide a cross-court permission rule.
Assess privacy and an unintended disclosure separately
The OAIC's commercial-AI guidance addresses personal-information handling and recommends due diligence, internal policies, training and ongoing review. Confirm which obligations apply to the firm and proposed use.
Map the destination AI provider's handling separately from the firm's security controls. Review access, retention, deletion and cross-border handling. A region or contract is not, by itself, a compliance finding.
If information is submitted outside the approved process, use the firm's incident procedure. Have the responsible legal/privacy owner assess containment, client communications and any notification obligations.
Evaluate supported controls without overstating coverage
Aona secures employees' AI use. Visibility requires the applicable browser plugin or native endpoint on the covered device. Catalogue recognition does not establish observation or enforcement for every tool.
Verify the application, account, browser or OS, input path, client release and configuration. Review the response and available event using synthetic data. Start with coverage and deployment requirements.
Backend hosting and Aona prompt processing are separate choices and do not determine the destination AI provider's handling. Aona evidence does not establish matter context, acknowledgement, privilege or legal compliance by itself.
Move from the worksheet to the firm's decision
Use the worksheet with the law-firm policy template, the product/settings review and the responsible lawyer's decision. Keep unresolved questions visible.
The legal-team solution is the next step for evaluating employee-AI visibility and supported data controls. Bring one rule, a synthetic input and the device/application scope.
Review after a material provider, client, policy or court-direction change. The worksheet supports an existing decision process; it is not an automated approval or matter-management system.
Sources and scope
Sources checked: 21 September 2026. This is an editorial summary of published Australian professional and privacy guidance, not legal advice. No professional legal review is recorded.
Primary guidance: joint regulators’ statement; ASCR resource; Law Society of NSW guide; OAIC guidance.
Have the responsible lawyer confirm the current rules, client instructions and court directions for the firm's intended AI uses. Outside Australia, consult applicable local professional guidance.
