AI and the Privacy Act: What Australian Real Estate Agencies Need to Know

Every AI tool an agency switches on touches personal information. A voice agent answers a buyer enquiry and records the call. An SMS assistant follows up an appraisal and logs the vendor's reply. AI doesn't change the kind of data you hold, but it changes where that data goes, who processes it, and what gets inferred from it. That is what AI privacy for real estate in Australia comes down to. Here are the parts of the Privacy Act 1988 that matter when an agency adopts AI, the amendments that start in December 2026, and the questions to put to any vendor before you hand over your database.

This article is general information, not legal advice. Talk to a privacy lawyer about your agency's specific situation.

First question: does the Privacy Act cover your agency?

The Act applies to organisations with an annual turnover of more than $3 million. The Office of the Australian Information Commissioner (OAIC) is clear that turnover here means all income from all sources, and that the test is whether you've exceeded $3 million in any financial year since 2002. A busy sales and rental office can cross that line sooner than the principal expects.

Below the threshold, the small business exemption is narrower than many agencies assume. Regardless of turnover, the OAIC lists businesses that are still covered, and several are relevant to real estate:

• Disclosing personal information for a benefit without consent (say, passing vendor details to a broker for a referral fee) counts as trading in it, and brings you in.
• A related body corporate of a covered company is covered.
• Residential tenancy database operators are covered.
• Reporting entities under the Anti-Money Laundering and Counter-Terrorism Financing Act 2006 are covered for their AML/CTF activities, whatever their size. That now includes most sales agencies: AUSTRAC regulates real estate professionals from 1 July 2026 when they broker the sale, purchase or transfer of real estate, for example as a buyer's or seller's agent. If your agency sells property, assume the APPs apply to the customer due diligence information you collect.
• A small business can voluntarily opt in.

The practical point: if your agency is covered, every AI tool you use has to fit within the Australian Privacy Principles (APPs). If you're genuinely exempt, the APPs are still the benchmark clients will judge you against. Build to them either way.

The five APPs that do the most work when you use AI

There are 13 APPs. For an agency running AI voice, SMS or writing tools, five of them carry most of the weight.

APP 1: open and transparent management. You need a clear, current privacy policy and the practices to back it up. If AI now handles enquiries or follow up, your policy should say so in plain language.

APP 3: collection. You can only collect personal information that is reasonably necessary, by lawful and fair means. Two AI traps sit here. A voice agent asking buyers questions you don't need answered is over-collecting. And the OAIC treats information that AI generates or infers about an identifiable person as a collection. An AI that tags a contact "likely divorcing" has created personal information you must justify, and some inferences stray into sensitive information that generally needs consent.

APP 6: use and disclosure. Information collected for one purpose stays with that purpose, unless the person consents or would reasonably expect a related secondary use. A vendor who booked an appraisal expects follow up about it. Feeding your whole database to a third party model for training is a different conversation.

APP 8: cross-border disclosure. Before disclosing personal information overseas, you must take reasonable steps to ensure the recipient won't breach the APPs, and under section 16C you are generally accountable for what that recipient does. Many AI products route data through offshore servers or model providers, so find out where your recordings and transcripts go.

APP 11: security. You must take reasonable steps to protect information from misuse, interference, loss and unauthorised access, and destroy or de-identify it once it's no longer needed. The APP now spells out that reasonable steps include technical and organisational measures. For AI, ask how long transcripts are kept and who at the vendor can read them.

The 2024 amendments and the 10 December 2026 deadline

The Privacy and Other Legislation Amendment Act 2024 followed the federal government's review of the Privacy Act. Many of its changes are already in force, including new penalty provisions and the APP 11 security wording above, which applies to information held from 11 December 2024.

The change most relevant to AI has a delayed start. From 10 December 2026, new APPs 1.7 to 1.9 require extra detail in your privacy policy if you've arranged for a computer program to make, or substantially and directly assist, a decision that could reasonably be expected to significantly affect someone's rights or interests using their personal information. The policy must describe the kinds of personal information used and the kinds of decisions involved.

The OAIC's updated APP 1 guidance reads the rule broadly. "Computer program" covers rule-based software and generative AI. A program can be in scope even when a human makes the final call, and even when it belongs to a vendor. One OAIC example is a rental company whose software scores applicants and drops the result into the assessment screen.

A voice agent that books appraisals is unlikely to be making decisions of that weight. For property management, any tool that scores, ranks or screens rental applicants deserves a hard look before December.

What the OAIC expects when you use off-the-shelf AI

In October 2024 the OAIC published guidance on privacy and the use of commercially available AI products, aimed at businesses buying AI rather than building it. The headline points:

• Privacy obligations apply to personal information you put into an AI system and to any personal information it produces.
• Do due diligence before you buy: testing for your use case, human oversight, security risks, and who can access your inputs.
• Update your privacy policy and notices, and clearly identify public facing AI tools as AI.
• As best practice, don't enter personal information into publicly available generative AI chatbots.
• Take a privacy by design approach, including a privacy impact assessment.

The fourth point catches a lot of agencies. Pasting vendor notes into a free chatbot to "write a nice follow up" is exactly what the OAIC warns against. A one page staff AI policy closes most of that gap.

Telling callers and SMS recipients they're dealing with AI

The OAIC says people should be made aware when they're interacting with AI rather than a human, and presenting an AI as a person could also raise misleading conduct issues under the Australian Consumer Law. Disclose early and plainly.

A voice agent opening:

"Hi Sarah, it's the AI assistant for Jordan at Harbourside Property, calling about Tuesday's appraisal. This call is recorded so Jordan can review it. Is now a good time?"

An SMS:

"Hi Sarah, this is Jordan's AI assistant at Harbourside Property. Thanks for coming through 14 Wattle St on Saturday. Want the contract of sale? Reply STOP to opt out."

Where AI sends marketing SMS, the Spam Act 2003 applies exactly as it would to a human. The ACMA requires consent, accurate sender identification with contact details, and a functional unsubscribe honoured within 5 working days. The ACMA also notes alphanumeric sender IDs generally can't receive replies, so "Reply STOP" needs a number that can. For outbound marketing calls, the Do Not Call Register and permitted calling hours still apply: weekdays 9am to 8pm, Saturday 9am to 5pm, no calls on Sundays or national public holidays. Our guide to SMS marketing laws for real estate in 2026 goes deeper on consent records.

An AI privacy checklist for real estate in Australia

Use this before you sign, and again at renewal.

Data location and flow

• Where are recordings, transcripts and contact data stored, and in which countries?
• Which model providers process our data? Is it used to train models for anyone else?

Collection and use

• Can we limit what the AI asks callers to what we need?
• Does it infer or tag anything about contacts beyond what they told us?
• Does it disclose that it's AI on every call and message?

Security and retention

• Who at the vendor can access our data, and is access logged?
• Can we set retention periods, delete on request, and get our data back if we cancel?
• How fast will they notify us of a data breach?

Decisions and oversight

• Does the product score, rank or screen people? Can staff override every output?

Messaging compliance

• Does every SMS identify our agency with a working opt-out, synced back to our CRM?
• Does it check the Do Not Call Register and block calls outside permitted hours?

If a vendor can't answer these in writing, that is your answer. When you assess Voqo, we'd expect you to ask every one of them.

Frequently Asked Questions

Is my agency exempt from the Privacy Act if turnover is under $3 million?

Possibly, but check. You're still covered if you trade in personal information without consent, are related to a covered body corporate, operate a residential tenancy database, are an AML/CTF reporting entity (which most sales agencies have been since 1 July 2026), or have opted in. The OAIC's small business checklist walks through each test.

Do I have to tell people they're speaking to an AI?

The OAIC's guidance says public facing AI tools should be clearly identified. Disclosing at the start of each call and message is the safe, simple standard.

Can I paste contact notes into ChatGPT to write follow ups?

The OAIC recommends, as best practice, not entering personal information into publicly available generative AI tools. Use a business tool with contractual controls over storage, access and training.

What changes on 10 December 2026?

Covered entities must describe in their privacy policy any use of computer programs, including AI, to make or substantially assist decisions that could significantly affect individuals. Rental applicant screening tools are the most likely trigger for agencies.

Does the Spam Act apply if an AI sends the SMS?

Yes. The Act regulates the commercial message, not who wrote it. You need consent, sender identification and a functional unsubscribe either way.

None of this is a reason to avoid AI. An AI agent that discloses itself, collects only what it needs and logs every conversation is easier to audit than a team texting from personal mobiles. If you're working out how AI voice for real estate fits your agency, start with the checklist above and choose the vendor that answers it without flinching.