The Privacy Act and AI tools: what Australian businesses need to know in 2026

The short version: Australia has no separate AI privacy law for businesses. The Privacy Act 1988 and its 13 Australian Privacy Principles (APPs) apply to AI the same way they apply to any other handling of personal information. The OAIC has said so explicitly, published specific guidance for commercially available AI products, and from 10 December 2026 new transparency rules cover decisions that computer programs, including generative AI, make or substantially inform.

This guide covers whether the Act applies to you, the principles that matter most when you use AI, what changed with the 2024 amendments, and a checklist to work through. It's part of our guide to sovereign AI for Australian operators, which covers where AI data travels in more detail. Everything here is general information, not legal advice. If a decision turns on it, talk to a privacy lawyer.

First question: does the Privacy Act apply to your business?

The OAIC's small business guidance says a small business is one with annual turnover of A$3 million or less, and many are exempt. Turnover here means income from all sources, not profit. But a long list of businesses are covered regardless of turnover, including:

• health service providers, which the OAIC reads broadly, including allied health professionals and complementary therapists
• businesses that trade in personal information
• contractors providing services under a Commonwealth contract
• operators of residential tenancy databases
• credit reporting bodies and reporting entities under the AML/CTF Act
• businesses related to a business the Act covers, and businesses that have opted in

Two practical notes. If you provide health-related services, such as allied health, counselling or behaviour support, assume you are probably covered and check; our piece on AI for NDIS practice administration looks at that world specifically. And even if you are exempt, your clients increasingly expect APP-level handling, and some of your larger customers may require it in contracts. Many exempt businesses treat the APPs as the standard anyway.

The Australian Privacy Principles that matter most for AI

All 13 APPs apply, but these seven do most of the work when you use AI tools. The summaries draw on the OAIC's APP quick reference and its AI products guidance.

APP 8 is the one that catches businesses using overseas AI services. We explain the use-versus-disclosure distinction and section 16C accountability in the sovereign AI guide.

What the OAIC's AI guidance tells businesses to do

In October 2024 the OAIC published guidance on privacy and the use of commercially available AI products, updated in January 2025. Its five headline takeaways, paraphrased:

1. Privacy obligations apply to personal information you put into an AI system and to output that contains personal information. Do due diligence before you adopt a product.
2. Update your privacy policy and notices to explain your AI use, and make clear when public-facing tools such as chatbots are AI.
3. Personal information an AI generates or infers is a collection under APP 3, including hallucinations about an identifiable person.
4. APP 6 limits using personal information in AI for purposes other than the one it was collected for, unless the person consented or would reasonably expect it.
5. As a matter of best practice, don't enter personal information, and particularly sensitive information, into publicly available generative AI tools.

The fifth point is where most small businesses are exposed. Staff pasting client emails into a free chatbot to 'tidy up a reply' is exactly what the OAIC is warning about. The fix is not to ban AI; it's to give people a tool you have vetted, with terms you understand, and a rule about what may go into it. Our AI security checklist is a starting point for the vetting.

What the 2024 amendments changed, with dates

The Privacy and Other Legislation Amendment Act 2024 received Royal Assent on 10 December 2024. It is the first tranche of reform; broader changes, such as a 'fair and reasonable' test for handling personal information, were left for later. Three parts matter most for businesses using AI.

Security: technical and organisational measures (in force)

APP 11 now states that reasonable security steps include technical and organisational measures. The OAIC's APP 11 guidelines describe technical measures as software, hardware and physical controls, and organisational measures as policies, procedures and staff training. For AI that means both the settings (scoped access, approval gates) and the rules (what staff may paste where) count.

The statutory tort for serious invasions of privacy (from 10 June 2025)

According to the OAIC's statutory tort page, a new cause of action for serious invasions of privacy commenced on 10 June 2025. Individuals bring it through the courts, not through the OAIC, and the OAIC says it extends to individuals and entities that are not necessarily APP entities, which can include small businesses exempt from the APPs. The threshold is a serious invasion, not an everyday slip, but it is one more reason not to treat the small business exemption as a free pass.

Automated decision transparency (from 10 December 2026)

New APP 1.7 to 1.9 commence on 10 December 2026. The OAIC's fact sheet (September 2026) says an APP entity is in scope where it has arranged for a computer program to make a decision, or do something substantially and directly related to making it; the decision could reasonably be expected to significantly affect an individual's rights or interests; and personal information is used in the process.

Three points in the OAIC's material stand out for small businesses using AI. 'Computer program' explicitly includes generative AI and chatbots. A decision can be in scope even where a human reviews the output. And the obligation generally stays with you, the business making the decision, even when the software comes from a vendor. If you are in scope, your privacy policy must describe the kinds of personal information used, the kinds of decisions made solely by computer programs, and the kinds where a program does something substantially related to the decision. The OAIC's advice if you're unsure: take a cautious approach and include the information. Its September 2026 announcement links the full resources.

What might be in scope for a small business? Using AI to shortlist job applicants, to decide whether a tenancy application goes forward, to prioritise which clients get services first, or to set individual pricing. What usually isn't: drafting a reply email you then edit and send, or summarising your own notes.

Sample privacy policy wording for AI use

Illustrative wording only. Adapt it to what you actually do, and have it checked if the transparency obligation may apply to you.

Notice what the template forces you to know: which tools, which data, where, and who approves. If you can't fill in the blanks, that's the work to do before 10 December 2026.

If something goes wrong: data breaches and AI

AI tools don't change the Notifiable Data Breaches scheme. Under the OAIC's NDB guidance, if you suspect an eligible data breach you must take reasonable steps to assess it within 30 days, and notify affected individuals and the OAIC as soon as practicable once you believe an eligible breach has occurred. An agent that emails the wrong client's records to the wrong person is a potential breach like any other. A good audit trail for AI agent actions is what lets you work out quickly what was exposed and to whom.

A Privacy Act checklist for AI tools

• Confirm whether the Act applies to you, including the turnover exceptions.
• List every AI tool in use, including the free ones staff use without asking.
• For each, record what personal information goes in, where it is processed, and whether it is retained or used for training.
• Set a written rule: no personal or sensitive information in public AI tools.
• Require human review before AI output is sent to a person or saved about them.
• Update your privacy policy and collection notices to cover AI use.
• Decide whether any AI-assisted decision could significantly affect someone, and if so prepare the APP 1.8 disclosures before 10 December 2026.
• Make sure agent actions are logged so you can investigate a suspected breach within 30 days.
• Review the list every six months, or when you add a tool.

Fulcrum Cortex is designed to make several of these easier: it is 100% Australian owned, offers sovereign AI models from SCX.ai for sensitive work, and is approval-first, so agents draft and you approve before anything goes out. It doesn't make you compliant on its own; that depends on how you use it. You can read how we handle data in our privacy policy.