Briefing
Reference 13 · Primer · About 5 minutes
Download the MP3 · Leonard, an AI voice, reads a script written from the sources below.
Written from the DTA's published policy, standard and plan, checked on 11 November 2026. Paraphrased, not reproduced. Check the current text before relying on it.
Before we record, every claim the hosts make is checked against the sources (this table was last updated 11 November 2026). Confirmed means a primary source or two independent reports; Reported means one party says so; Partly right, Disputed, Unverified and Not yet known mean what they say, and the hosts say so on air. This is the table as we left it.
The rules that apply to AI in Australia and for Australian organisations with European exposure, and the practical work each one calls for.
These are Agile Insights AI Strategy & Governance offerings. Which of them you need depends on where your organisation starts.
Sign up for our podcast at agile-insights.com.au.
Leonard On the fifteenth of December 2026, every Commonwealth agency hits its biggest AI deadline yet. Three documents set the rules behind it. By the end of this episode, you will know which are mandatory and what each one asks. You will also know what changes for suppliers.
Leonard First, what they are. The Digital Transformation Agency, the D T A, publishes two of them. The Policy for the responsible use of AI in government is now at version two. It took effect on the fifteenth of December 2025, and it is mandatory for non corporate Commonwealth entities. The AI technical standard came out in July 2025. The policy strongly recommends it, but does not make it mandatory. The third document is the AI Plan for the Australian Public Service, announced in November 2025.
Leonard The problem these solve is the gap between enthusiasm and accountability. An agency can buy a chatbot in a week. Working out who owns it, whether anyone assessed it, and how it will be switched off takes far longer. Picture an internal audit asking how many AI tools a department runs. Without a register, nobody can answer, and nobody is accountable for the answer.
Leonard Who uses them? Agency heads and chief AI officers use the policy to set direction and name owners. Architects and delivery teams use the technical standard as a build and run checklist. Procurement teams use both when they write contracts. Suppliers meet them in tender questions, and in clauses about logging, monitoring and transparency.
Leonard So what is in them? Start with the policy. It runs in phases. Agencies had to publish a strategic position on AI by June 2026. By the fifteenth of December 2026, they need a register of AI use cases. Each one needs a named owner. Staff need foundational AI training, and impact assessments must be under way. Use cases already running must be assessed by the thirtieth of April 2027.
Leonard Next, the technical standard. It follows a system through three phases: discover, operate and retire. Within those sit eight stages, from design and data through to monitoring and decommissioning. Each statement carries criteria marked required or recommended. Here is the subtlety. The standard itself is not mandatory. But an agency that claims to meet it must meet every required criterion. In June 2026 the D T A added an addendum for AI agents. It asks that authorised staff can watch an agent in real time, step in, and pause or stop it.
Leonard Then the plan. The AI Plan for the public service is the adoption half, where the policy is the guardrails half. It rests on three pillars: trust, people and tools. It brings chief AI officers, foundational training, and a government AI platform called GovAI. For suppliers, two items matter most. There are expectations of external service providers, and a central register of generative AI assessments.
Leonard How does this connect to the rest of the stack? All three point back to Australia's AI Ethics Principles, which are voluntary on their own. The policy makes them bite inside government. The D T A's AI Reference Architecture, covered in Reference 12, traces the standard's requirements into system design. Security sits beside all of this, under the protective security framework, and that is Reference 14.
Leonard Getting started. First, build the register now, with an owner for every use case, including tools that staff already use. Second, list the use cases that are already running, because the April 2027 date covers them. Third, if you are a supplier, map your product to the technical standard's required criteria before a tender asks. The primary sources are the policy, the standard and the plan, on the digital dot gov dot A U website. They are linked in the show notes.
Leonard Now, currency. As at November 2026, the policy is at version two, unchanged since it took effect. The technical standard dates from July 2025, with the agentic addendum from June 2026. Check the current text before you rely on any detail here.
Leonard In short, the policy is the rule, the standard is the yardstick, and the plan is the push to adopt. If you'd like to know more, sign up for our podcast at agile-insights.com.au.
All of the reference series (16) · Follow AI Governance Reference in Apple Podcasts or any podcast app
Everything from Agile Insights: briefings, Making You Think and the reference series