Australia Defence Policy Settings for Responsible Use of AI

Australia’s live Defence governance page is titled Artificial intelligence. The policy it points to is the Policy Settings for Responsible Use of Artificial Intelligence in Defence. Official pages still say Artificial Intelligence and AI. This article uses those names. Super Intelligence (SI) is this site’s own-voice term for current U.S. executive-branch policy; it does not rewrite an Australian instrument. The NATO Principles of Responsible Use remain an Allied baseline, not an Australian contract mark.

A separate 2021 DST technical report, A Method for Ethical AI in Defence, is research. It is not the Policy Settings.

What the Policy Settings page actually lists

Defence says AI technologies offer opportunities to drive greater accuracy, efficiency, speed and safety in Defence. Defence will use AI responsibly and in compliance with Australia’s domestic laws and international legal obligations. It will apply an informed, risk-based approach, which will preserve individual accountability for AI-enabled decisions and outcomes.

The Policy Settings for Responsible Use of AI in Defence outline Defence’s approach by meeting the following obligations:

1. Lawfulness. Defence will use all technologies in compliance with Australia’s domestic laws and international legal obligations.
2. Adherence to values-based principles. Defence will ensure use of AI is underpinned by individual accountability; bounded by consideration of impacts on humans; explainable; reliable and secure; and designed to mitigate unintended bias and unintentional harm.
3. Proportionate controls. Technology use will be risk-based. Risks will be managed through proportionate control measures.

The same page lists a PDF of the Policy Settings for Responsible Use of Artificial Intelligence in Defence and a contact for [email protected]. We did not retrieve a stable defence.gov.au file URL for that PDF on this pass, so we quote the HTML rather than a mirror.

The Australian Army Research Centre dates the Policy Settings as released by the Department of Defence (Defence) in March 2026. That Army Research Centre page is a legal-review essay, not the Policy Settings themselves. It is cited here only for that March 2026 date.

The values-based list on the Defence HTML is five items, not a product certificate: individual accountability; bounded by consideration of impacts on humans; explainable; reliable and secure; and designed to mitigate unintended bias and unintentional harm. Proportionate controls remain risk-based with proportionate control measures.

No page we cite treats those settings as a CMMC analogue, a NATO PRU certificate, or UK MOD JSP 936. Completing the US AIA Toolkit does not, on any page we cite, satisfy them.

What DST’s 2021 method is, and is not

DST’s A Method for Ethical AI in Defence records a workshop held from 30 July to 1 August 2019. The PDF says the report articulates the views of participants and outcomes of the workshop for further consideration and does not represent the views of the Australian Government.

The publication page groups workshop topics into five facets: Responsibility – who is responsible for AI?; Governance – how is AI controlled?; Trust – how can AI be trusted?; Law – how can AI be used lawfully?; and Traceability – How are the actions of AI recorded?. The method includes three tools: an Ethical AI for Defence Checklist, Ethical AI Risk Matrix and a Legal and Ethical Assurance Program Plan (LEAPP).

Treat those tools as workshop outputs that will be provided to support the development of Defence policy, doctrine, research and project management. They are not the later Policy Settings. The PDF’s disclaimer is the status line: it articulates the views of participants and outcomes of the workshop for further consideration and does not represent the views of the Australian Government. A 2019 workshop method that the publisher says is not government policy is not a 2026 Policy Settings substitute.

See also Canada DND/CAF Artificial Intelligence Strategy and NATO’s revised AI Strategy.

Questions

Are the Policy Settings a product certificate for suppliers?

No source we cite treats them as a product or supplier certificate. They are Defence’s published obligations for how Defence will use AI.

Does the 2019 workshop method bind Defence contracts?

The DST PDF says the report does not represent the views of the Australian Government. We found no cited page that makes the checklist, risk matrix or LEAPP a contract clause.