Start with applicability and purpose
Confirm the entity's coverage, the activity and the kinds of information involved. Identify the intended purpose and how it relates to the organisation's existing information practices.
Check whether the new AI task fits the purpose for which the existing information was collected. Consent is one ground for a new use, and APP 6 sets out others.
Follow information through the system
Document inputs, retrieval sources, prompts, generated outputs, storage, integrations and logs. Identify vendors and how each one handles the information.
Then establish:
- who can access the information
- what is retained
- whether it can be used for another purpose
- which locations and recipients matter.
Treat unknown vendor settings as open questions to resolve.
Examine quality and effects on people
AI can produce inaccurate information about an individual or derive new inferences from existing records. Find out how quality is checked for the purpose, and how errors will be found and dealt with.
Consider the practical effect of the output, including whether it enters a significant decision. A low-cost tool can still create consequential information practices.
Review safeguards and operation
Access, confidentiality, security, deletion and retention need to reflect the information and use. Appropriate arrangements may include limiting inputs, restricting permissions, training users and evaluating outputs.
A staff policy is one control. It does not show that every use follows it.
Treat automated-decision disclosure separately
From 10 December 2026, APP 1.7 and APP 1.8 set privacy policy disclosure rules for some automated decisions. They sit alongside the rest of the privacy assessment.
Prepare a decision record
Record the facts, assessment, specialist advice, allowed boundary, controls, owner and reassessment triggers. Where a material fact remains unknown, state what must be resolved before proceeding.
This page is general guidance. It is not legal advice or a complete compliance determination.
Frequently asked questions
What should a privacy review of an AI workflow cover?
It should cover every point where personal information enters, moves through or leaves the AI workflow, with the controls at each point. The Office of the Australian Information Commissioner's guidance on commercially available AI products is a useful checklist for it.
Does every AI use need consent under the Privacy Act?
Not every use needs consent. Under APP 6, a use the individual would reasonably expect, and that is related to the original purpose, can proceed without it. For sensitive information, the use must be directly related.
Is a staff AI policy enough for privacy compliance?
A staff AI policy helps, and the review still needs evidence from the workflow itself. Check access settings, retention settings and a sample of real prompts and outputs.
Does APP 1.7 apply to every AI activity?
APP 1.7 applies only where all of its conditions are met, and some uses it reaches are not AI at all. The APP 1.7 scope and disclosure guide sets out the conditions.
APP 1.7 disclosure applies from 10 December 2026. A Rapid Review maps the relevant decisions and the disclosure work still required.