← Australian context
Australian context

The Australian Privacy Principles and AI, in plain English

What the Privacy Act means when the data is employee and pay data and the tool is an AI assistant.

As of November 2026
How this page was checked. The Office of the Australian Information Commissioner (OAIC) is the privacy regulator. Every statement about the law on this page was read from an OAIC page in November 2026 and links to it. We describe what the OAIC says. We do not tell you what the law requires of your organisation: that is a question for your privacy officer.

What the APPs are and who they cover

The Privacy Act 1988 is the main federal privacy law. The Australian Privacy Principles (APPs) sit inside it. They set the rules for how personal information is collected, used, stored and disclosed.

There are 13 APPs. The OAIC lists them, with a one-line purpose for each, in its APP quick reference. Checked November 2026

Who is covered. The OAIC says Australian Government agencies, and organisations with an annual turnover of more than $3 million, have responsibilities under the Privacy Act, subject to some exceptions. Checked November 2026

  • Some smaller businesses are covered too. The OAIC's list includes private sector health service providers, businesses that sell or buy personal information, contracted service providers for an Australian Government contract, and businesses that have opted in.
  • State and territory government agencies are not covered by the Privacy Act. The OAIC says they come under state and territory legislation.
  • The OAIC's list of what is not covered includes, "in some situations", the handling of employee records by an organisation. That is the exemption explained in the next section.

Source: OAIC, rights and responsibilities. Whether your own organisation is covered is a question for your privacy officer.

The employee records exemption

An exemption is a carve-out from the rules. The OAIC describes an "employee records exemption" for private sector employers. Checked November 2026

  • It applies only in limited circumstances: handling that is directly related to a current or former employment relationship, and that involves an employee record.
  • The OAIC's examples of what an employee record covers include salary or wages, performance or conduct, leave, and trade union membership.
  • It does not cover government agencies. Their employee records are covered by the APPs.
  • It does not cover future employment relationships, so job candidates are outside it.
  • It does not cover contractors who handle another organisation's employee information.
  • It does not cover volunteers.

Sources: OAIC, employee records exemption and OAIC, employment.

Is it still there? In November 2026 the OAIC's rights and responsibilities page still listed the handling of employee records by an organisation, in some situations, among the things the Privacy Act does not cover. Checked November 2026 Removing the exemption has been discussed as part of privacy reform, so ask your privacy officer for the current position.

The practical point, in our words: even where the exemption may apply, pay data is sensitive in every practical sense. Contracts and other laws may still apply. A sensible default is to handle pay data as if the APPs applied.

The APPs that matter most with AI

These six come up most when an AI assistant touches people data. The first sentence of each is the OAIC's own summary of the principle. The second is our note on why it matters for AI.

  • APP 1, open and transparent management of personal information. The OAIC says it ensures personal information is managed in an open and transparent way, including through a clearly expressed and up to date privacy policy. Our note: if AI becomes part of how your organisation handles people data, the privacy policy may need to say so.
  • APP 3, collection of solicited personal information. The OAIC says it outlines when personal information can be collected, and applies higher standards to sensitive information. Our note: the OAIC's AI guidance treats information an AI system generates or infers about a person as a collection.
  • APP 5, notification of the collection of personal information. The OAIC says it outlines when and in what circumstances a person must be told about certain matters. Our note: think about what employees have been told about how their information is used.
  • APP 6, use or disclosure of personal information. The OAIC says it outlines the circumstances in which personal information that is held may be used or disclosed. Our note: this is the one to think about before putting data collected for payroll into an AI tool for a different purpose.
  • APP 8, cross-border disclosure of personal information. The OAIC says it outlines the steps to take to protect personal information before it is disclosed overseas. Our note: ask where an AI service processes and stores data.
  • APP 11, security of personal information. The OAIC says reasonable steps must be taken to protect personal information from misuse, interference and loss, and from unauthorised access, modification or disclosure. Our note: ask who can see what goes into the tool, and what comes out.

Checked November 2026 Source for the summaries: OAIC, APP quick reference. The notes are ours, not the OAIC's.

What the OAIC says about AI

The OAIC publishes guidance on privacy and the use of commercially available AI products, meaning tools you can buy or sign up for rather than build. It also publishes guidance on developing and training AI models.

The guidance on commercially available AI products was published on 21 October 2024 and updated on 17 January 2025. It also covers freely available tools, such as public AI chatbots. Its main points, in the OAIC's terms: Checked November 2026

  • The Privacy Act applies to all uses of AI involving personal information. Privacy obligations apply to personal information put into an AI system, and to what the AI produces where that contains personal information.
  • Before adopting a product, do due diligence: whether it is suitable and has been tested for the intended use, how human oversight will work, the privacy and security risks, and who will have access to the information.
  • Update privacy policies and notifications with clear information about the use of AI, and set policies and procedures for using AI systems.
  • Information an AI generates or infers about an identifiable person is personal information, even when it is wrong. Generating it counts as a collection under APP 3.
  • APP 6 applies when personal information is put into an AI system: it should be used for the primary purpose it was collected for, unless there is consent or the person would reasonably expect the secondary use and it is related to the primary purpose.
  • The OAIC recommends a Privacy Impact Assessment (a structured review of privacy risks before you start) as part of a "privacy by design" approach.

The line most relevant to this site: "As a matter of best practice, the OAIC recommends that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools".

Source: OAIC, guidance on privacy and the use of commercially available AI products. It is long and worth reading in full, including its checklists.

Recent and coming changes

The Privacy and Other Legislation Amendment Act 2024 changed the Privacy Act. One change concerns automated decisions, where a computer program makes or helps make a decision using personal information.

Starting 10 December 2026. The OAIC says that from this date, organisations and agencies that use personal information in automated decision making with the potential to affect a person's rights or interests will be required to say so in their privacy policies: the kinds of personal information used, and the kinds of decisions made this way. Checked November 2026

The OAIC consulted on guidance for this obligation in mid-2026. Source: OAIC, consultation on guidance for transparency in automated decision making. We did not find the final guidance, so check the OAIC site for it.

Why this matters here, in our words: a pay recommendation or job sizing tool that substantially shapes a decision about a person is the kind of thing to raise with your privacy officer before you rely on it. Every use case on this site keeps a person making the decision.

Still under discussion. Further reform has been proposed, including to the small business and employee records exemptions. We have not confirmed where those proposals stand, so do not assume either way. Ask your privacy officer.

"Sensitive information"

There are two meanings, and they are not the same.

  • The legal meaning. The OAIC describes sensitive information as personal information about a person's racial or ethnic origin, political opinions or associations, religious or philosophical beliefs, trade union membership or associations, sexual orientation or practices, criminal record, health or genetic information, and some aspects of biometric information. It generally has a higher level of privacy protection. Checked November 2026
  • The everyday meaning. Information that would cause harm or embarrassment if it leaked. Pay data is sensitive in this sense.

Pay is not on the OAIC's list. So a salary is personal information when it can be linked to a person, but it is not "sensitive information" in the legal sense. Some things HR holds are on the list, such as health information and union membership, and they often sit in the same files.

The OAIC also lists "employee record information" as an example of personal information. Source: OAIC, what is personal information? In practice, treat pay data with great care whatever its legal label.

What this means for the use cases on this site

Each use case touches different data. This table is our guide, not the OAIC's.

Use caseData it touchesThe habit that keeps it safe
01 Pay gap analyserA workforce file, processed in your browserUse IDs, not names
02 Job evaluationRole descriptionsDescribe roles, not people
03 Pay recommendationRole, level and pay positionsPositions, not names
04 Conversation coachChoices in a practice conversationChoices only; no names in practice
05 Team assistantPolicy documentsRestrict who has access
06 Remuneration analyserReal pay dataOnly inside your own approved environment
07 Manager's review trackerA team's pay, ratings and written reasons, saved to a fileIDs not names; save the file where confidential documents belong; delete working copies when the review closes
08 Reward team's review trackerThe whole review's pay data, saved to a fileTreat the saved file like a remuneration committee paper; restrict the folder to the reward team
09 Remuneration health checkIndividual pay, ratings and reporting lines, read on your own computerNothing is saved or sent; share the grouped summary, keep the list of individuals as working data

Six habits

These are our advice, not the OAIC's.

  1. Invented data first. Build and test with made-up data. Move to real data only when the method works.
  2. IDs, not names. Replace names with codes before data goes near a tool.
  3. Aggregate before you share. Share averages and ranges, not individual rows.
  4. Only employer-approved tools for real data. If your employer has not approved a tool, keep real data out of it.
  5. Hide small groups. A group of two or three people is easy to identify. Suppress or combine them.
  6. A person checks before anything is sent or decided. AI drafts. A person reviews and owns the outcome.

Questions for your privacy officer

  • Do the APPs apply to our organisation, and does the employee records exemption apply to what I want to do?
  • Which AI tools are approved for personal information, and for pay data?
  • Where does each approved tool process and store data, including overseas?
  • Is pay information treated as sensitive under our policies?
  • Does our privacy policy need to change for the automated decision rules, and from when?
  • Who signs off before I use real data in a new tool?

To match the tool to the data, see Which tool for which data.