Rowland to draft AI code for public service automated decisions

Digital Attitudes, 21 July 2026

Term 3 is running everywhere by this morning. Three stories from the last seven days that land on the same Term 3 agenda: a federal code of practice on automated decisions, a K-12 literacy program that pulled Google.org money without waiting for a state signal, and a Sydney industry consortium putting research anchors under the graduate outcomes question.

Rowland's AI code catches up with automated decisions across the public service

Attorney-General Michelle Rowland will develop a code of practice for automated decision-making that applies across the whole federal public service, The Mandarin reported on Monday 20 July. The initiative sits inside the AI safety priorities Prime Minister Anthony Albanese announced at the University of Sydney last Tuesday, and it names the specific failure mode: agencies increasingly rely on AI for administrative decisions without transparency or human oversight in the chain. Industry Minister Tim Ayres is on the same brief. Rowland's framing line: "there's always a person and responsibility in the chain of decision making, where it's providing better and fairer services for Australians" (The Mandarin on public service AI guardrails, 20 July).

Why this matters: For every state education department running automated at-risk flagging, welfare-trigger escalations, attendance nudging or timetable optimisation, the federal code will set the vocabulary the states have to mirror, because the state DoE is a public service body and the vendor conversation will move faster once a Commonwealth code names an implementation date. ACARA and Education Services Australia are also caught by it, which means the "actions to mitigate platform interruptions to future NAPLAN test events" flagged in the 15 July ministers' communique now has a second document to align to. For a Commonwealth-funded university (ANU is the clearest case, but the CRC and ARC linkages pull others in), the code writes itself into the Notifiable Data Breach and DPIA processes on the same page as the OAIC Children's Online Privacy Code. The concrete Term 3 action for a state DoE, TAFE or Commonwealth-funded body is to have the "which of our AI-in-admin decisions falls under this code" register drafted before Rowland publishes an implementation date, because vendor contract terms will move quickly after that.

Day of AI wins Australia's only slot in Google.org's GenAI Accelerator

Day of AI Australia and UNSW confirmed on Friday 17 July that the partnership has been selected as the sole Australian participant in Google.org's Generative AI Accelerator, joining a global cohort of 20 organisations that share a $30 million pool plus six months of pro bono technical support from Google's AI specialists. The funding and the training go into an evergreen year-round version of "Win the Farm" (the program's existing generative AI game), a new ethics game for the 2026 cohort, and an expanded device donation program for under-resourced schools. Since launch in 2022, Day of AI Australia has reached more than 100,000 students nationwide (Education Matters on the Australian partnership named in the Google.org GenAI Accelerator, 17 July).

Why this matters: For every Australian primary and secondary leader trying to work out how to move on AI literacy this term without waiting for a state or federal framework, Day of AI has just cleared the funding runway for the next 18 months of curriculum-adjacent K-12 material without a Department of Education line item. The ethics game is the specific piece worth requesting a preview of before the Term 3 professional practice day, because a game-based ethics module lands in classrooms where a policy document does not, and it will arrive before any state authority produces its own aligned resource. For an independent or Catholic sector CIO, the pattern (external Big Tech funding routed through a UNSW-anchored non-profit rather than direct vendor procurement) is the template that sidesteps the AFT-style gold-standard question that hit Google when it partnered with the Utah State Board last month. The other read is upstream: the Office of AI in Prime Minister and Cabinet now has a live K-12 AI literacy exemplar in the field, and any national K-12 AI standard drafted under the National AI Plan will have to explain why it doesn't cite Day of AI's material.

Sydney signs four sector anchors to a three-year AI-and-work consortium

The University of Sydney announced on Tuesday 14 July a three-year industry consortium with Ashurst Perkins Coie, Commonwealth Bank of Australia, Deloitte and Telstra to research how AI will change Australian professions. Sydney Executive Plus co-directors Associate Professor Sandra Peter and Professor Kai Riemer lead an interdisciplinary team drawing on information systems, innovation, organisational studies and law, examining leadership competencies in AI-driven workplaces, workforce upskilling, work design, productivity, employee experience and professional identity. Vice-Chancellor Professor Mark Scott framed the initiative as "research for good" with outputs that are "practical, productive and people-first", with potential international expansion after year-one pilots (University of Sydney on collaborating with industry to make AI work, 14 July).

Why this matters: For every DVC (Academic) and DVC (Education) writing the 2027 graduate outcomes brief, Sydney has just handed the sector the closest thing yet to an industry-signed research agenda on what an AI-transformed profession actually looks like, and the four names on the consortium (a Big Four accountant, a top-tier law firm, a Big Four bank, a national telco) are the same firms whose graduate hiring signals ripple into TEQSA-audited program reviews across the country. Expect Sydney to publish an "AI-transformed professions" framework in year one, and any university whose own graduate outcomes framework does not reference it will get the question at the next Academic Board. The cross-disciplinary framing (law and information systems and innovation and organisation studies on one project) is also the direct answer to the Turnitin data point from earlier this month: 10.2 per cent AI-heavy Australian submissions is a symptom, the assessment redesign work at CQU and UTS is a partial fix, but the graduate-outcome side of the equation has not had a research anchor until this week. For a DVC (Research) pitching to a state-based industry cluster, the Sydney template (three-year commitment, four sector-anchor firms, one academic lead pair) is a lot cheaper to replicate than a single-firm partnership.

Term 3 is running. Rules for the code, funding for the classroom, research for the graduate. Coffee's on. More at digitalattitudes.com.au.