Privacy Act duties when a law firm uses AI
A law firm covered by the Privacy Act 1988 must apply the Australian Privacy Principles to what it puts into an AI tool and what the tool produces. From 10 December 2026 its privacy policy must also describe decisions that computer programs make or substantially inform. This guide covers the principles that matter most, the exemption and the statutory tort.
It is general information about the law as at October 2026, not legal advice. Yes AI is not a law firm.
Is your firm covered by the Privacy Act?
The Australian Privacy Principles (APPs) apply to Australian Government agencies and to private sector organisations, but not to every business. The OAIC’s small business page explains that a business with an annual turnover of 3 million dollars or less is generally exempt, unless it falls into a group covered regardless of size. That group includes health service providers, businesses that trade in personal information, contractors providing services under a Commonwealth contract, businesses related to a covered business, and businesses that have opted in.
Plenty of small practices sit under the threshold. Being exempt does not remove the other duties. Rule 9 of the conduct rules still governs client information (see AI and solicitors’ professional obligations), clients expect the same care either way, and the OAIC suggests small businesses consider opting in to build trust. Firms that act for government under Commonwealth contracts should check that exception carefully.
Which privacy principles matter most for AI?
Legal matters often involve what the Act calls sensitive information. The OAIC’s definition includes criminal record, health or genetic information, sexual orientation, religious beliefs and racial or ethnic origin, all common in criminal, family, injury and employment files. Sensitive information gets a higher level of protection.
| Principle | What it requires (summary from the OAIC quick reference) | How it applies to AI in a law firm |
|---|---|---|
| APP 1 | Manage personal information openly, with a clear and current privacy policy | Say in the policy that AI answers calls, transcribes or drafts, and from 10 December 2026 describe automated decisions |
| APP 3 | Rules for collecting solicited information, with higher standards for sensitive information | An intake agent should ask only for what the firm needs before the conflict check |
| APP 5 | Tell people about certain matters when you collect their information | Callers hear that they are speaking with AI and whether the call is recorded or transcribed |
| APP 6 | Use or disclose information only for the purpose it was collected, unless an exception applies | Pasting a client’s file into a new AI tool for an unrelated purpose needs a lawful basis |
| APP 8 | Take steps to protect information before it is disclosed overseas | Find out where the AI vendor processes and stores data |
| APP 10 | Take reasonable steps to keep information accurate, up to date and complete | Transcription and summary errors need checking before they are relied on |
| APP 11 | Protect information from misuse, loss and unauthorised access, and destroy or de-identify it when no longer needed | Set retention periods for recordings, transcripts and prompts |
| APP 12 | Give people access to information held about them | A caller may ask for their message record or a recording |
What did the OAIC say about commercially available AI products?
On 21 October 2024 the OAIC published guidance on privacy and the use of commercially available AI products, aimed at organisations that deploy AI tools such as chatbots, transcription and writing assistants. Its main points, in plain terms:
- Privacy obligations apply to any personal information entered into an AI system, and to output that contains personal information.
- If an AI system generates or infers personal information, that is a collection, and the collection rules apply.
- Under APP 6, information should be used for the purpose it was collected, unless a secondary use is within the person’s reasonable expectations or another exception applies.
- As a matter of best practice, organisations should not enter personal information, and particularly sensitive information, into publicly available generative AI tools.
- Due diligence on an AI product is not “set and forget”; it continues while the product is in use, and the OAIC recommends a privacy by design approach that includes a privacy impact assessment.
What changes on 10 December 2026?
The Privacy and Other Legislation Amendment Act 2024 added new transparency requirements to APP 1. The OAIC’s announcement of its guidance, released on 30 September 2026, confirms that from 10 December 2026 an APP entity that has arranged for a computer program to make, or substantially and directly inform, a decision that could significantly affect someone’s rights or interests must include three things in its privacy policy:
- the kinds of personal information used in running those programs
- the kinds of decisions made solely by those programs
- the kinds of decisions where the programs do something substantially and directly related to making the decision.
Most AI use in a law firm is drafting, transcription or message-taking, where a person makes every decision. The tools to look at closely are those that decide something about a person on their own: for example, an intake system that automatically declines enquiries in practice areas the firm does not handle, or one that decides which enquirers get a call back. List every such tool, decide with your advisers whether it is in scope, and update the policy before the date.
Where does your AI vendor process data?
APP 8 requires reasonable steps before personal information is disclosed overseas. The OAIC’s APP 8 guidelines explain that the Australian entity can be accountable for what the overseas recipient does with the information. Giving information to an overseas cloud provider is generally a disclosure, though it may be treated as a use where the provider is contractually limited to storage and the firm keeps effective control.
Many AI products process data in more than one country. Yes AI’s own privacy policy says information may be processed outside Australia, including for cloud hosting and AI platform services, and that client projects may use different hosting, communications and AI providers. For a law firm, ask any vendor, Yes AI included, for a written list of the providers that handle call audio, transcripts and messages, the countries involved, and the retention periods. Where a provider offers an Australian data region, ask whether it can be used. The private AI page and customllm.au cover options that keep processing under tighter control.
What is the statutory tort for serious invasions of privacy?
The same 2024 amending Act added a statutory tort to the Privacy Act. According to the OAIC’s page on the tort, it started on 10 June 2025 and covers two kinds of conduct: intruding on a person’s seclusion, and misusing information that relates to them, where the person had a reasonable expectation of privacy. Claims go to the courts; the OAIC has no direct role in administering the tort. For a law firm, the practical lesson is that recordings and transcripts of distressed callers need the same care as the rest of the client file.
How common are data breaches in professional firms?
The OAIC’s Notifiable Data Breaches figures for 2025, released on 6 July 2026, recorded 1,205 notifications for the year, with 716 caused by malicious or criminal attacks. Legal, accounting and management services made 81 notifications. Every AI tool that holds client data is one more place a breach can start, which is why the vendor list and retention settings matter.
A short action list
- Confirm whether the Privacy Act covers your firm, and record the answer.
- List every AI tool in use, what personal information goes into it, and where the vendor processes it.
- Update the privacy policy to describe AI use, and the automated decision content before 10 December 2026.
- Add an AI and recording notice to the phone greeting and the web enquiry form.
- Set retention periods for recordings, transcripts, prompts and messages.
- Put the rules into the firm’s AI policy using the AI use policy checklist.
To discuss how a Yes AI agent would handle client information at your firm, send an enquiry without including details of any legal matter.
Frequently asked questions
- Does the small business exemption mean a small firm can ignore privacy when using AI?
No. A firm under the turnover threshold may be outside the Australian Privacy Principles, but solicitors still owe clients confidentiality under the conduct rules, and clients expect their information to be handled carefully whatever the firm's size. The OAIC also encourages small businesses to consider opting in to the Privacy Act.
- Is an AI transcript of a phone call personal information?
Usually, yes. A transcript or summary that identifies the caller, or is about a person who can be identified, is personal information. The OAIC's guidance on commercial AI products treats information generated or inferred by an AI system as a collection, so the collection, use and security principles apply to it.
- Do we need a privacy impact assessment before using an AI agent?
The OAIC recommends a privacy by design approach that includes a privacy impact assessment when introducing an AI product. For a reception or intake agent, a short assessment covering what is collected, who sees it, where it is processed and how long it is kept is a sensible minimum.
- Which AI tools could be caught by the December 2026 change?
Tools that make, or substantially and directly inform, decisions that could significantly affect someone's rights or interests, using personal information. Drafting aids and message-taking, where a person decides, are less likely to be in scope than a system that decides on its own which enquiries are declined. Check each tool with your advisers.