Australian Government AI Policy and Frameworks¶

Purpose: Understand the Australian Government's approach to AI governance — policy, guidance, institutions and timelines Audience: Governance, compliance, risk and strategy teams | Time: 30-40 minutes

Australia does not yet have a standalone AI Act. On 15 July 2026, Prime Minister Albanese announced plans to legislate an Australian Standards for AI framework for large data centres and AI training, including mandatory requirements for large AI data centres — see Australian Standards for AI and Office of AI below. The announcement is not itself legislation. In the interim, the Government relies on existing technology-neutral laws (see Current Legal Landscape) supported by voluntary guidance, a growing set of institutional arrangements and — for Commonwealth agencies — existing mandatory AI policy requirements.

This page covers government policy, guidance frameworks and institutional developments. For the laws themselves, see Current Legal Landscape for AI in Australia.

Australian Standards for AI and Office of AI (July 2026)¶

What was announced:

• Australian Standards for AI: The Government plans to legislate a national framework for large data centres and AI training. National Cabinet met on 26 August 2026 and endorsed developing the standards as mandatory, with the Commonwealth committing to legislate them, in consultation with states and territories, in early 2027. The final scope, duties and commencement arrangements have not yet been published, and the 26 August outcome is pending confirmation against a primary government source — see National Cabinet outcome below.

• Office of AI (established 15 July 2026): The Office sits within the Department of the Prime Minister and Cabinet. PM&C says it will coordinate across Australian Government agencies to design and legislate the new standard, working with the responsible ministers and the Department of Industry, Science and Resources.

• Proposed data centre requirements: The announced standards would require large AI data centres to underwrite new power supply, pay their share of connection costs, reduce power when needed to support the grid and be as water-efficient as possible. These proposals build on the existing Data Centre Expectations (March 2026), but require legislation before they become legal obligations.

• Copyright: The announcement says Australian writers, artists and journalists should retain ownership over their work and that companies should not use Australian creative works to train AI without the artist's control. It does not specify the legal mechanism. See Current Legal Landscape for the separate April 2026 TDM development.

Relationship to AI6 and VAISS: The announcement does not withdraw AI6 (Guidance for AI Adoption, October 2025) or the VAISS guardrails. They remain published, non-binding guidance. Organisations must continue to comply with existing laws while the proposed standards are designed and considered.

On 20 July 2026, the Government separately announced five whole-of-government AI safety priorities: a Digital Duty of Care, further privacy reform, workplace AI safety, examination of consumer-law options, and a framework for automated decision-making in federal agencies. These are policy workstreams rather than enacted duties. (AI consumer safety priorities, accessed 22 July 2026)

Sources: Prime Minister of Australia, Office of AI — PM&C, Department of Industry ministers (accessed 22 July 2026)

National Cabinet endorses mandatory standards (26 August 2026)¶

National Cabinet met on 26 August 2026 and agreed that large data centres have material impacts on energy, water and land use. It endorsed developing nationally consistent mandatory standards covering those areas, plus skills and training requirements, and committed the Commonwealth to work with states and territories to legislate the standards in early 2027. This is scoped to data-centre infrastructure — it does not itself extend the AI6 organisational-governance guidance or introduce mandatory guardrails for general AI use.

Joint Select Committee on Artificial Intelligence (submissions close 14 September 2026)¶

A bipartisan federal Joint Select Committee on Artificial Intelligence was appointed by resolution of the House of Representatives and the Senate on 20 August 2026, with government and Coalition backing. It will examine the adequacy of existing AI-related laws and regulatory frameworks — including copyright and intellectual property law, national security implications, data sovereignty, consumer protection, deepfakes and cyber security — and the rate and depth of AI adoption among Australian businesses, with a particular focus on small, medium and family businesses.

Submissions close 14 September 2026 and the committee is due to report by 30 November 2026. This is a new, ongoing avenue of federal AI scrutiny, separate from the Senate Select Committee on Adopting AI (whose government response is covered above) and the Senate's Artificial Intelligence and Data Centres inquiry (see AI Data Centres and Compute Governance).

(Sources: Liberal Party of Australia, SmartCompany, accessed 30 August 2026; Parliament of Australia, accessed 6 September 2026)

National AI Plan (December 2025)¶

The National AI Plan sets Australia's national roadmap for AI. Key positions:

• Australia will rely on existing laws and sector regulators rather than introducing a standalone AI Act or immediate mandatory guardrails
• A new AI Safety Institute will monitor, test and advise on emerging AI risks
• Government investment in AI infrastructure, skills and research through specific programs (AI Adopt Centres, AI Accelerator CRC, APS AI Plan)

The Government's formal response to the Senate Select Committee on Adopting AI was tabled on 1 April 2026, confirming these positions and committing to:

• Acceptance in principle of mandatory guardrails for high-risk AI contexts, with further consultation on implementation
• Sector regulators as the primary enforcement mechanism (APRA, ASIC, ACCC, OAIC) — no new standalone AI regulator
• The AI Safety Institute will provide guidance and assessments (not enforcement)
• Launch of the AI Accelerator CRC (CRC-P Round 19 closed 12 May 2026; CRC Round 28 expected 2027)
• Rejection of a text-and-data-mining exception in copyright law; the Copyright and AI Reference Group (CAIRG) is the consultation vehicle (see legislation page)
• Consultation on AI's impact on the creative sector through the next National Cultural Policy process
• Affirmation of the Privacy Act automated decision-making reforms (effective 10 December 2026)

Full response (industry.gov.au).

AI Safety Institute¶

The Australian AI Safety Institute (AISI) was formally established in early 2026 with $29.9 million in committed funding. As of July 2026, the AISI is actively testing unreleased frontier AI models — the most substantive public operational milestone for the institute to date.

Leadership:

• Dr Kate Conroy (philosopher, RAAF reservist) — inaugural General Manager, appointed May 2026
• Professor Paul Salmon — safety science research lead, appointed July 2026

Remit:

• Technical assessments of advanced AI systems, including testing of unreleased frontier models
• Bilateral and multilateral engagement through the International Network of AI Safety Institutes
• Publishing research on AI safety

Key developments (as of July 2026):

• At the AI Safety Forum (University of Sydney, 7–8 July 2026), Assistant Minister Charlton warned that frontier AI models "are already doing things their creators never intended: cheating, deceiving, going their own way", framing the AISI's testing lab as the appropriate place to catch deceptive AI behaviour before deployment
• The AISI is testing unreleased frontier models in collaboration with the Australian Signals Directorate (ASD) as technical partner
• Bilateral AI safety agreements with the UK and Canadian AI safety institutes were signed in May 2026, giving the AISI access to shared testing methods and risk intelligence
• Two active research projects announced at the AI Safety Forum:
  • With the Gradient Institute: assessing the risk of AI agents acting autonomously on behalf of humans
  • With CSIRO: investigating human oversight and verification of capable AI systems at scale (supported by UK partners)
• The AISI signed a Memorandum of Understanding with Anthropic on 1 April 2026 — the first formal industry collaboration under the National AI Plan. The MOU covers AI safety evaluation, infrastructure alignment and capability building. It is non-binding and does not confer preferential treatment in procurement
• On 9 August 2026, the AISI published its first substantive report, "Risks and controls for multi-agent systems", commissioned from the Gradient Institute — the output of the research project noted above. It sets out a technical framework mapping risks, failure types and controls across three deployment tiers (singular governance, federated governance and open environments) for situations where AI agents from different organisations interact, such as a business's agent transacting with a customer's or supplier's agent. Coverage describes it as the first government publication to treat cross-organisational AI agent risk as a distinct governance problem

(Sources: minister.industry.gov.au, innovationaus.com, startupdaily.net, accessed 12 July 2026; industry.gov.au, gradientinstitute.org, accessed 23 August 2026)

The AISI advises; portfolio agencies and sector regulators retain enforcement responsibility.

DTA Mandatory AI Requirements (Commonwealth Agencies)¶

The DTA updated its Policy for the Responsible Use of AI in Government (effective 15 December 2025). Key elements:

• Mandatory AI Impact Assessments using a new AI Impact Assessment Tool (12-section fillable template aligned with AI Ethics Principles)
• Mandatory AI procurement guidance
• Mandatory foundational AI training for all Australian Public Service staff
• Requirement for agencies to appoint Chief AI Officers

Timeline:

While these are mandatory only for Commonwealth agencies, they signal the direction for all Australian organisations. The DTA's AI Impact Assessment Tool is a practical resource for any organisation assessing AI risk.

GovAI Chat — an AI assistant for APS staff — began trials from April 2026 as part of the broader APS AI Plan.

Data Centre and AI Infrastructure Expectations (March 2026)¶

The Department of Industry, Science and Resources published voluntary-but-consequential expectations for data centre and AI infrastructure developers on 23 March 2026.

Key expectations:

• Fair and well-paid Australian jobs
• Investment in domestic workforce development
• Contribution to research and innovation
• Underwriting new renewable power supply

Data centre proposals not closely aligned with these expectations will not be prioritised by Commonwealth regulatory assessments, giving the document practical regulatory weight.

Guidance for AI Adoption (AI6)¶

The Guidance for AI Adoption (October 2025) is the primary government guidance for responsible AI governance and adoption. Published by the National AI Centre (NAIC), it sets out six essential practices:

1. Decide accountability
2. Understand impacts
3. Measure and manage risks
4. Share information
5. Test and monitor
6. Maintain human control

Two versions are available: Foundations (for organisations starting out) and Implementation practices (for more mature organisations). Supporting tools include an AI screening tool, AI policy guide and template, and AI register template.

For detailed coverage, see Guidance for AI Adoption (AI6).

Voluntary AI Safety Standard (VAISS)¶

The Voluntary AI Safety Standard (September 2024) introduced 10 guardrails for safe and responsible AI. These guardrails remain relevant as a detailed control set and have been fully integrated into the Guidance for AI Adoption through an official "VAISS × implementation practices" crosswalk.

For detailed coverage of the 10 guardrails, see Voluntary AI Safety Standard.

Government–Industry Arrangements¶

ACCC Enforcement Approach¶

The ACCC published its own AI transparency statement in February 2026 and has flagged "AI-washing" (misleading claims about AI capabilities) as an enforcement concern. The Treasury's review of AI and the Australian Consumer Law found the existing framework "fit for purpose", making dedicated AI consumer legislation unlikely in the near term.

Key Dates¶

Key References¶

• National AI Plan — industry.gov.au
• Office of AI — pmc.gov.au
• AI consumer safety priorities — minister.industry.gov.au
• AI Safety Institute — industry.gov.au
• Risks and controls for multi-agent systems — industry.gov.au
• DTA AI Policy Update — dta.gov.au
• AI Impact Assessment Tool — digital.gov.au
• Government Response to Senate Select Committee on AI — industry.gov.au
• Data Centre Expectations — industry.gov.au
• Guidance for AI Adoption (AI6) — industry.gov.au
• Voluntary AI Safety Standard — industry.gov.au