One citizen-facing automated decision, walked through seven surfaces and scored against the Australian AI / ADM / privacy obligation mesh. The picture below is the readiness diagnostic — compliant, exposed, gap, or not-yet-known — against the 10 December 2026 ADM transparency deadline.
01 · S5 Notification
AI involvement isn't disclosed
breaches G1 (APP 1.7–1.9) · G4
Good looks like: a live disclosure in the privacy policy and at the point of decision, with reasons adequate for an ART merits review — in place before 10 Dec 2026.
02 · S2 Data matching
No screening for proxy bias
risks G5 (Rec 17.1) · G9
Good looks like: matching logic documented and tested, low-confidence matches routed to a human, and proxy-variable screening for indirect discrimination on the record.
03 · S4 The human decision
Rubber-stamp, not a decision
risks G2 · G4 · G10
Good looks like: defined review gates, overrides logged with reasons, and a named accountable decision-maker who is the person who actually reviews the output.
Readiness on 10 December 2026
Is this decision on track to be disclosable, explainable and reviewable?
Not yet. S5 disclosure is not live, and S2/S4 carry unresolved fairness and accountability exposure. Two surfaces are unknown — evidence simply isn't held — which is itself a finding.
AT RISK