The Digital Transformation Agency has switched on a tougher version of the Commonwealth’s AI rulebook, requiring every mainstream federal agency to name accountable officials, keep a register of every AI use case and put all staff through mandatory training as the technology spreads through the public service.
The updated Policy for the responsible use of AI in government, known as Version 2.0, took effect on 15 December 2025. It replaces the lighter-touch first version that had governed Commonwealth AI since September 2024.
The DTA set out the changes in an article titled “AI Policy Update: Strengthening responsible use across government”. The full rules sit on the government’s digital.gov.au policy pages.
The policy applies to all non-corporate Commonwealth entities, with carve-outs for the defence portfolio and the national intelligence community. Corporate Commonwealth entities are encouraged, but not required, to follow it.
What agencies now have to do
The core shift is from broad principles to hard, dated obligations. Agencies must develop a strategic approach to adopting AI, stand up governance to operationalise responsible use, and assign clear accountability for each AI use case.
Every agency must keep an internal register of all in-scope use cases, with a named accountable owner for each one. Each agency must also identify accountable officials and report them to the DTA within 90 days of the policy taking effect.
Within six months, agencies must publish a public transparency statement setting out how they adopt and use AI. Foundational AI training becomes mandatory for all staff across the Australian Public Service, building a consistent baseline of understanding.
The requirements are being phased in rather than dropped all at once. The DTA says the first new mandatory requirement begins on 15 June 2026, with the remaining obligations coming into force in December 2026.
The agency is holding itself to the same standard. According to the DTA’s own transparency statement, it has designated two accountable officials: AI Branch Manager Ramsey Beydoun, responsible for cross-government coordination, and Chief Operating Officer Tom Gilmartin, responsible for the DTA’s own implementation plan.
Building on a policy already in force
Version 2.0 tightens a framework that has been running for more than a year. The first policy took effect on 1 September 2024 and introduced the original duties to appoint accountable officials and publish transparency statements.
Take-up of that first round has been near-universal. In an update on its central register, the DTA reported that all 94 agencies subject to the transparency standard had met their publishing obligations, with a further 20 agencies volunteering statements.
The new rules land alongside a broader push announced weeks earlier. The APS AI Plan 2025 was released on 12 November 2025 by Finance Minister Katy Gallagher.
Reporting on the rollout, The Modern Regulator noted the plan mandates foundational AI literacy for the roughly 200,000 people across more than 150 agencies who make up the public service, alongside a secure APS-wide chatbot cleared to operate at the PROTECTED classification level.
Announcing the plan, Gallagher framed AI adoption as a matter of public confidence. “Trust is our licence to operate,” she said, adding that the shift was about “unlocking new capabilities” rather than replacing people, according to The Modern Regulator.
Unions want guardrails for workers, not just systems
The public sector unions have given the training push a guarded welcome while pressing for firmer protections. The Community and Public Sector Union backed the training commitment but warned that consultation with staff “is not optional”, according to The Modern Regulator.
The Australian Council of Trade Unions has gone further, calling for mandatory AI Implementation Agreements, job guarantees and a new National AI Authority to oversee the rollout across the wider economy.
The tension is familiar. The government is presenting AI as a productivity tool for stretched agencies, while worker representatives want binding commitments on jobs and consultation before systems are switched on at scale.
Why it matters for Australia
For Australians, this is about how the systems that decide payments, visas, tax and service eligibility are built and checked. Requiring a named owner for every use case, plus a public transparency statement, gives citizens and oversight bodies a clearer line of sight into where automated tools touch their dealings with government.
It also reshapes work for the public service itself. Mandatory training for every APS employee is one of the largest workforce upskilling exercises the Commonwealth has attempted, and it sets a benchmark that state governments and large private employers will watch closely.
There is a sovereignty dimension too. By standing up a PROTECTED-level chatbot and insisting on internal registers, the Commonwealth is trying to keep sensitive data inside controlled systems rather than pushing officials toward consumer AI tools it cannot govern.
The economic stakes are large. The Modern Regulator cited estimates that Australia risks forgoing a $142 billion AI opportunity without regulatory clarity, and reported that most local organisations still provide no AI training at all, leaving the public service as an unusually early mover on baseline literacy.
What comes next
The immediate test is the 15 June 2026 milestone, when the first new mandatory requirement takes effect, followed by the full suite of obligations in December 2026. Agencies now have a countdown to register their use cases, lock in accountable officials and finish training their staff.
The bigger question sits above the public service. Industry Minister Tim Ayres is expected to release a national AI plan covering the wider economy, and the unions’ call for a National AI Authority and mandatory guardrails remains unresolved.
For now, Canberra has chosen to lead by governing its own use of AI first. Whether that discipline extends to how it regulates AI in the rest of the economy is the debate still to come.
Sources: Digital Transformation Agency – AI Policy Update: Strengthening responsible use across government; digital.gov.au – Policy for the responsible use of AI in government (Version 2.0); Digital Transformation Agency – AI transparency statement; Digital Transformation Agency – New central register of AI transparency statements; digital.gov.au – APS AI Plan 2025; The Modern Regulator – How Australia’s public service is rolling out its AI training mandate.
