Albanese puts AI office inside PM&C to centralise policy

By Zara Kincaid · 2026-07-26

New office, new address

The Office of AI was established inside the Department of the Prime Minister and Cabinet on 15 July 2026verifiedVerified Source: pm.gov.au, placing AI policy at the apex of the federal bureaucracy rather than inside a line agency.^[1] Prime Minister Anthony Albanese said the move was deliberate. "This world-leading framework is about Australia choosing to shape the future rather than letting the future of AI shape us," Albanese said.^[1]

The office's mandate goes beyond coordination. It will design and legislate new Australian AI standards and unlock AI training across the public and private sectors.^[1]

What the framework actually covers

The Office of AI will enforce mandatory requirements for large AI data centres to underwrite their own power supply, pay full network connection costs, reduce power when the grid demands it and minimise water use.verifiedVerified Source: pm.gov.au^[1] Those infrastructure rules sit alongside a wider policy agenda covering a digital duty of care and direct action on the risks that chatbots pose to children.^[2]

The framework also spans AI's intersection with skills, manufacturing, defence and national security.^[2] Albanese offered no itemised funding breakdown for individual workstreams at the time of announcement.

The regulatory foundation already in place

The government released its Expectations of data centres and AI infrastructure developers on 23 March 2026, setting mandatory energy, water and social licence requirements as the foundation for the new standards the office will legislate.verifiedVerified Source: industry.gov.au^[5] That document predates the office by nearly four months, meaning the new body inherits a regulatory base to draft binding law from rather than starting with a blank sheet.

The National AI Plan, launched in December 2025, had already established a broader governance architecture bringing together government, industry, research and community bodies, and created the National AI Centre to guide adoption by small businesses and agencies.

Anthropic MOU and the Microsoft deal

On 1 April 2026, the Australian Government signed a non-legally binding Memorandum of Understanding with Anthropic to support the National AI Plan.^[3] Anthropic committed to aligning any future Australian operations with the government's data-centre expectations, expanding its local presence, supporting Australian research and academic institutions, and collaborating with the Australian Public Service on capability building.^[3]

The MOU carries no enforcement mechanism. Its commitments are prospective, applying only to future Australian operations rather than requiring immediate structural changes from Anthropic.^[3]

The Digital Transformation Agency moved earlier and harder on procurement. On 26 February 2026, the DTA secured a five-year Volume Sourcing Agreement with Microsoft, commencing 1 July 2026, locking in Commonwealth agency access to Microsoft 365, Copilot and Azure at capped pricing with enhanced legal protections.^[4] That agreement predates the Office of AI's establishment by more than four months, meaning the government's largest AI procurement commitment was made before the coordinating body existed.

Industry and critics respond

The Australian Computer Society welcomed the announcement but signalled the hardest work lies ahead. "ACS hopes the Office of AI will provide the coordination needed to connect policy, standards, workforce planning, and implementation across government and industry," ACS CEO Dr Prins Ralston said.^[7] Ralston's statement positioned the office as a necessary but still-unproven mechanism for joining up initiatives that have so far run in parallel.

Electronic Frontiers Australia offered a more pointed assessment. EFA Chair John Pane said his organisation backed the shift away from voluntary guidelines but had clear reservations. "We are supportive of the proposed transition away from voluntary safety and security guidelines but are deeply skeptical about the lack of immediate enforcement details and whether they will be as strong as those currently in place in the EU, which are human rights preserving," Pane said.^[6]

EFA's concern centres on the gap between said intent and operational readiness. The EU's AI Act carries fines of up to 35 million euros or seven per cent of global turnover for the most serious violations; no equivalent penalty architecture has been published for the Australian standards the Office of AI is tasked with writing.^[6]

What remains unanswered

Privacy implications of the framework have not been detailed in the government's published materials. How the Office of AI will interact with the Office of the Australian Information Commissioner on data-handling obligations for AI systems remains an open question.

The DTA's Microsoft Volume Sourcing Agreement, signed on 26 February 2026 and active from 1 July 2026, means Commonwealth agencies are already deploying Copilot under a procurement framework designed before the new office's standards exist.^[4]

Zara Kincaid

Zara Kincaid covers breaking news and sport for Bushletter. Fast and verb-led, he writes with a news-wire cadence and no patience for PR spin.