Framework overview
AustraliaAlignment note · 2 Oct 2026

How OCC formats line up with Australian guidance

Australia already has the principles. The Open Conformance Coalition (OCC) offers small, open, royalty-free formats that make some of them checkable in practice. This page maps those formats to the national guidance. It is an alignment note, not a claim of endorsement or of compliance by any agency.

National AI Centre, Oct 2025

Guidance for AI Adoption (AI6)

Digital Transformation Agency, v2.0 (from 15 Dec 2025)

Policy for the responsible use of AI in government

Agents specifically

Agentic AI guidance and the new reporting standard

Five Eyes — Careful adoption of agentic AI services (ASD, CISA, NSA, NCSC and partners)
May 2026

Know what each agent is and what it can reach; least privilege; logging; human oversight; the ability to stop it.

Agent record identity, mandate limits, evidence and revocation
Australian AI Safety Institute — Risks and controls for multi-agent systems
Aug 2026

Risks that emerge when agents coordinate, delegate and share capabilities.

T-RUE delegation and revocation; AIH-12 H6 Networked
Immediate-reporting standard for rogue AI incidents (announced)
29 Sep 2026

Report immediately to the affected organisation and to authorities; legislation intended by end of 2026.

Where to report (disclosure contacts) and in what form (incident record). OCC does not calculate deadlines.

Summaries are ours. Read the originals at cyber.gov.au and the Department of Industry, Science and Resources.

Privacy Act, from 10 December 2026

Automated decisions in privacy policies

APP 1.7–1.9 require privacy policies to describe the kinds of personal information used, and the kinds of decisions made solely or substantially by a computer program, where those decisions could significantly affect people. The Agent Record has optional adm fields that point to that disclosure for each agent. They do not replace the privacy policy.

Automated decisions & transparency fields
2026 landscape

Who is doing what

  • Office of AI, reported as established within PM&C in July 2026, coordinating whole-of-government AI policy.
  • Australian AI Safety Institute, advisory, established in early 2026; its multi-agent report is mapped above.
  • Joint Select Committee on AI, due to report by 30 November 2026. OCC’s submission sets out the asks below.
  • Data-centre standards, announced in July 2026 and reported as endorsed by National Cabinet in August. Out of scope for OCC.

Summaries are ours and based on public reporting; check the originals before relying on them.

PM&C, after the Medicare breach

Rapid review terms of reference

AreaWhat open formats can contribute
Reporting requirementsA structured incident record with aware_on and initial_report_on, and OECD-aligned harm coding.
Information sharingPublished disclosure contacts (/occ/disclosure-contacts.json) and open schemas any agency can adopt.
Obligations of AI firmsAn agent record an operator publishes for each agent: who runs it, its mandate, and how to revoke it.
Enforcement and deterrenceOut of scope. These are matters for government; OCC records can serve as evidence of what was declared and when.
Resilience of government systemsLargely out of scope; ASD and ACSC guidance applies. A disclosure contact gives each system a route for reports.

Source: pmc.gov.au. See the Medicare case study for the timeline.

What we are asking for
  1. When the reporting standard says how fast, also say where (a published disclosure contact) and in what form (a structured incident record). OCC’s drafts are one input; any open equivalent would serve.
  2. One bounded pilot with an agency or university, with failures published alongside successes.
  3. Reviewers who will tell us where the drafts are wrong. Review is not endorsement.