Automated decisions & transparency
From 10 December 2026, Australian organisations covered by the Privacy Act must say in their privacy policy when a computer program makes, or substantially shapes, decisions that could significantly affect people. In the EU, Article 50 transparency duties have applied since 2 August 2026. The OCC Agent Record now has optional fields that point to those disclosures, agent by agent, in an open format.
Australia: automated decisions in privacy policies
The duty applies where a program uses personal information to make a decision, or to do something substantially and directly related to making it, and the decision could reasonably be expected to significantly affect an individual’s rights or interests. It applies to decisions made on or after 10 December 2026, including where the personal information was collected earlier. Our summary; the Act and OAIC guidance govern.
| What the privacy policy must cover (summary) | Agent Record field |
|---|---|
| The kinds of personal information used in the operation of the computer program | adm.personal_information_kinds |
| The kinds of decisions made solely by the operation of the computer program | adm.solely_automated_decision_types |
| The kinds of decisions for which a thing substantially and directly related to making the decision is done by the program | adm.substantially_related_decision_types |
| Where the disclosure is made: the entity’s privacy policy | adm.privacy_policy |
Sources: OAIC — APP Guidelines, Chapter 1; OAIC automated decision-making guidance.
EU: telling people they are dealing with AI
| Article 50 duty (summary) | Agent Record field |
|---|---|
| Art. 50(1): people are informed they are interacting with an AI system, unless obvious from context | eu_art50.ai_interaction_disclosure |
| Art. 50(2): synthetic audio, image, video or text is marked in a machine-readable format and detectable as generated | eu_art50.synthetic_content_marking |
Under the Digital Omnibus amendments, systems already on the market have until 2 December 2026 to apply machine-readable marking. Our summary; the Regulation governs.
The fields
"adm": {
"solely_automated_decision_types": ["Approving refunds under AUD 200"],
"substantially_related_decision_types": ["Ranking loan applications for a credit officer"],
"personal_information_kinds": ["Transaction history", "Contact details"],
"privacy_policy": "https://example.org/privacy#automated-decisions"
},
"eu_art50": {
"ai_interaction_disclosure": "disclosed at start of interaction",
"synthetic_content_marking": "C2PA manifest on generated images"
}- All fields are optional. Leave them out if the agent makes no decisions that significantly affect people.
- Describe kinds of decisions and information, as the law does. Never put personal information in a record.
- Operators fill these in on the portal (step 7, “Automated decisions”). They appear on the public record and in record.json, signed with the rest of the record.
One place per agent
A privacy policy speaks for a whole organisation. An agent record speaks for one agent: who runs it, its mandate, its authority and how to stop it. Linking the two lets a reader see which agent makes which kinds of decisions, and lets the operator keep the record and the policy consistent when either changes.