Australia Privacy Act AI Compliance Checklist: Ten Items Before a Prompt Carries Personal Information
A working checklist for Australian organisations running AI over personal information under the Privacy Act 1988 and the Australian Privacy Principles, with the automated decision-making transparency obligations landing 10 December 2026. Each item names the obligation, the concrete action, and the evidence to produce, so the list doubles as an audit-readiness pass rather than a set of intentions. Ordered by what an OAIC inquiry reaches for first.
The Privacy Act 1988 (Cth) and the Australian Privacy Principles (APPs) already govern every AI system that touches personal information, and the Office of the Australian Information Commissioner (OAIC) enforces them without waiting for a dedicated AI Act. The Privacy and Other Legislation Amendment Act 2024 adds automated decision-making transparency obligations from 10 December 2026 and gives the OAIC infringement-notice and civil-penalty powers. This checklist is ordered the way an inquiry moves: identity and records first, then transfers, then decisions, then the paperwork.
Each item pairs the action with the evidence it produces, because a checklist that lists intentions without artifacts fails the moment someone asks you to prove it.
1. Bind corporate identity to every AI request
Route AI traffic so the named user or agent, not a shared service key, is attached to each prompt. Evidence: a per-request record showing which corporate identity sent which prompt to which model. Without this, APP 11 security claims rest on an application log that lost the identity upstream.
2. Classify what goes into the prompt
Evaluate the data classification of the context window before it leaves. Evidence: a per-request classification decision showing what category of personal information was permitted for that caller's role, satisfying APP 6's purpose-limitation on use.
3. Record the model destination and its region
Log the model endpoint and hosting region for every request carrying personal information. Evidence: a destination log that answers APP 8's cross-border disclosure question directly, rather than a quarterly vendor inventory that cannot show which prompt went overseas.
4. Enforce destination policy for sensitive classifications
Block or reroute prompts carrying higher-sensitivity personal information when the target endpoint sits in a disallowed region. Evidence comes from the blocked-request record itself. This is where AI data residency controls turn APP 8 from a statement into an enforced boundary.
5. Replace static API keys with per-role, per-route policy
Remove shared credentials that grant permanent full model access. Evidence: an access policy showing which roles can reach which models and data scopes, plus the enforcement log. Static keys are the most common APP 11 failure in AI deployments.
6. Produce a per-decision record for automated decisions
For any AI output that significantly affects an individual, write a record binding the input prompt to the output and the policy applied. Evidence takes the form of decision lineage across the request. This is the operational backing for the 10 December 2026 transparency obligation and mirrors GDPR Article 22 handling.
7. Flag high-impact decisions for human review
Route decisions with significant effect so a person can be inserted before the outcome is final. Evidence: the review-flag record showing which decisions were eligible for human review and whether it occurred.
8. Keep the audit record independent of the application
Ensure the compliance record is written by a system separate from the one making the AI call, so it survives selective logging, suppression, and crash-loss. Evidence: a signed, tamper-evident log outside application custody. The reasoning is in You Own the AI Liability, Not the Vendor.
9. Update the privacy policy for automated decisions
Before 10 December 2026, revise the APP 1 privacy policy to describe automated decisions that significantly affect individuals and the kinds of personal information used. Evidence: the published policy text, dated, with the per-decision record from item 6 standing behind the claim.
10. Pre-build the breach reconstruction path
Confirm you can reconstruct an AI-channel exposure end to end for the Notifiable Data Breaches scheme. Evidence: a query against the independent request log that produces a timeline. My honest take: this is the item teams skip and the one that decides how an incident actually goes.
How the checklist maps to the controls
Items 1 through 10 resolve to a small number of enforcement points at the request boundary, laid out in the Australian Privacy Principles controls mapping. The audit-evidence view covers what each artifact looks like when the OAIC asks for it. Together they turn this list from a set of tasks into a running system with evidence attached.
DeepInspect
Nine of these ten items produce evidence at the same place: the boundary between your users or agents and the LLM APIs they call. DeepInspect sits inline at exactly that boundary. It binds identity to each request, classifies the prompt, enforces destination and role policy, flags decisions for review, and writes a signed per-decision record before the response returns. The checklist becomes a query, not a scramble.
For an Australian organisation facing the December 2026 automated decision-making obligations, that means the privacy-policy update in item 9 rests on the decision lineage in item 6. Book a technical deep dive at deepinspect.ai.
Frequently asked questions
- Is this checklist enough for OAIC compliance?
It covers the AI-traffic obligations that an OAIC inquiry reaches for first: identity, security, cross-border disclosure, automated decisions, and breach reconstruction. It does not replace a full privacy program covering collection notices, retention, and data subject access across all channels. Treat it as the AI-specific layer that sits on top of an existing Privacy Act program.
- What has to be done before 10 December 2026 specifically?
Items 6, 7, and 9: the per-decision record, the human-review flag for significant decisions, and the privacy-policy update describing automated decisions. The transparency obligation is the surface requirement, and the OAIC's later inquiry into a complaint will test whether a real decision record stands behind the disclosure.
- Do small organisations have to follow this?
The Privacy Act's small-business exemption is narrowing under the reforms, and many AI-using organisations fall outside it because of the kinds of data they handle. Any entity bound by the APPs that runs AI over personal information needs these controls. If you are unsure of your status, treat the checklist as the safer default.