Weekly Digest
Week of 15 Dec 2025
This week at a glance
This week's most significant development for APS AI governance practitioners is the Department of Finance's publication of guidance establishing the Chief AI Officer role across the APS, with agencies required to appoint a CAIO from existing senior leadership by July 2026. Alongside this, Finance Secretary Matt Yannopoulos announced two supporting structures — AIDE, a whole-of-APS coordination function, and GovAI, a secure platform for AI collaboration and capability development including a generative AI tool planned for 2026 — marking a deliberate policy shift from agency-level experimentation toward coordinated, Finance-led adoption. Practitioners working on agency AI governance structures, role definitions, or implementation planning under the APS AI Plan will find this week's Finance guidance directly relevant to near-term decisions. At the international level, NIST's preliminary draft cybersecurity framework profile for AI (NISTIR 8596) is open for public comment until 30 January 2026 and may be a useful reference for agencies incorporating AI considerations into their existing cybersecurity and risk management approaches.
Headlines
- AU Gov · Fri 19 Dec 2025 Establishing Chief AI Officers for the APS Government Finance (Department), Finance (Portfolio)
- Standards · Draft NIST Guidelines Rethink Cybersecurity for the AI Era
- Practice · How we’re enabling research with sensitive data on AI supercomputers
- Risk · Towards Risk-Aware Artificial Intelligence and Machine Learning Systems: An Overview
Australian Government3 items
Fri 19 Dec 2025 Establishing Chief AI Officers for the APS Government Finance (Department), Finance (Portfolio)
The Department of Finance has published initial guidance establishing the Chief AI Officer (CAIO) role across the APS, with agencies required to appoint a CAIO from their existing senior leadership by July 2026. The CAIO role is distinct from the existing AI Accountable Official role: while Accountable Officials oversee governance and responsible adoption, CAIOs are expected to lead AI transformation, identify opportunities, and drive cultural change. Guidance notes flexibility - the CAIO may be combined with a CIO or CDO role, or held by a policy or operational leader, depending on the agency's context. A new AIDE function will coordinate CAIOs across the service.
Key points
- The APS AI Plan requires all agencies to appoint a Chief AI Officer (CAIO) from existing senior leadership by July 2026.
- CAIOs are distinct from AI Accountable Officials - they lead AI transformation and cultural change, not just governance.
- A new AI Delivery and Enablement (AIDE) function will coordinate CAIOs across the APS to drive safe AI adoption.
Implications
- Decide Agencies will need to decide who will fulfil the CAIO role and how it sits alongside existing AI Accountable Official and CIO/CDO arrangements before July 2026.
- Implement Senior leadership and governance teams should download the CAIO information pack and begin internal deliberations on role design and appointment processes.
- Monitor Agencies may want to monitor further guidance from the AIDE function as the CAIO framework matures alongside evolving AI best practice.
Thu 18 Dec 2025 AIDE and GovAI: moving from experimentation to impact across the APS Government Finance (Department)
The Secretary of the Department of Finance has announced the establishment of AI Delivery and Enablement (AIDE), a new whole-of-APS function to move government AI adoption from isolated experimentation to coordinated, scalable impact. AIDE will focus on shared adoption barriers, priority use cases, governance navigation, and cross-agency lesson sharing. Alongside AIDE, the GovAI platform - an APS-only environment for training, collaboration, and AI tool access - is being expanded, with GovAI Chat (secure generative AI for all APS staff) planned for 2026. Both initiatives are framed as operationalising the AI Plan for the APS.
Key points
- Finance Secretary announces AIDE - a new whole-of-APS function to drive coordinated, scalable AI adoption across government.
- GovAI platform provides APS-only secure AI collaboration and training, with GovAI Chat generative AI capability planned for 2026.
- The announcement signals a formal shift from agency-level pilots to system-wide, Finance-led AI uplift and enablement.
Implications
- Consider Agencies currently running or planning AI pilots may want to consider engaging with AIDE to share experiences, reduce duplication, and access system-wide support.
- Implement APS staff and AI leads should direct staff to GovAI's interactive training as an immediate, endorsed capability uplift pathway.
- Monitor Agencies could monitor the GovAI Chat rollout in 2026, as it will provide the first whole-of-government secure generative AI capability and will reshape agency tooling decisions.
Tue 16 Dec 2025 Celebrating innovation at GovHack 2025 Government Finance (Department)
The Department of Finance sponsored GovHack 2025, the southern hemisphere's largest open data hackathon, setting challenges focused on improving government service access and reducing regulatory overlap. Two winning teams - Team 2918 and BinChicken LLC - developed AI-powered platforms addressing service navigation and small business compliance. Finance framed the event as consistent with the APS AI Plan's collaboration and knowledge-sharing ambitions. Winning teams are invited to present at AI CoLab in 2026.
Key points
- Department of Finance sponsored GovHack 2025, setting challenges around government services and regulatory simplification.
- Winning teams developed AI-powered platforms to improve service navigation and small business compliance.
- Winners are invited to showcase at AI CoLab next year - limited direct governance or policy signal for APS practitioners.
Implications
- Monitor Agencies with an interest in AI-powered service delivery or regulatory simplification may want to watch what emerges from the AI CoLab showcase of winning GovHack concepts.
Standards & Frameworks1 item
Draft NIST Guidelines Rethink Cybersecurity for the AI Era
NIST has published a preliminary draft of its Cybersecurity Framework Profile for Artificial Intelligence (NISTIR 8596), designed to help organisations apply CSF 2.0 to AI-related cybersecurity risks and opportunities. The profile covers three focus areas: securing AI systems, using AI for cyber defence, and building resilience against AI-enabled threats. It was developed over a year with input from more than 6,500 stakeholders and is now open for public comment until 30 January 2026, with a workshop planned for 14 January 2026. A refined initial public draft is expected in 2026, after which it will include mappings to NIST's AI Risk Management Framework.
Key points
- NIST has released a preliminary draft Cyber AI Profile (NISTIR 8596) for 45-day public comment, closing 30 January 2026.
- The profile maps cybersecurity guidance across three areas: securing AI systems, AI-enabled defence, and AI-enabled attack resilience.
- Still in preliminary draft stage; a refined initial public draft is planned for 2026, limiting immediate applicability for Australian agencies.
Implications
- Monitor Agencies and ACSC-aligned security teams may want to monitor NISTIR 8596's development, given the likely influence of NIST frameworks on Australian cybersecurity and AI governance guidance.
- Consider APS AI and cybersecurity policy practitioners could consider whether submitting comment or attending the 14 January workshop aligns with their agency's engagement priorities before the 30 January deadline.
Public Sector Practice & Guidance1 item
How we’re enabling research with sensitive data on AI supercomputers
The Alan Turing Institute's FRIDGE project aims to enable researchers to safely use sensitive data on high-performance AI supercomputing infrastructure. The blog post frames this as unlocking frontier AI benefits without compromising data security or privacy. The extracted text is limited, so the specific technical or governance mechanisms employed are not assessable from this item. The challenge it addresses—reconciling sensitive data governance with access to large-scale compute—is relevant to Australian research agencies and public sector data custodians navigating similar constraints.
Key points
- The Alan Turing Institute's FRIDGE project enables secure research using sensitive data on AI supercomputers.
- Addresses a genuine governance challenge—safely accessing frontier compute for sensitive-data research—relevant to Australian research and public sector contexts.
- Extracted text is minimal; substantive detail of the approach is not available from this item alone.
Implications
- Monitor Australian agencies involved in sensitive-data research infrastructure (e.g. CSIRO, ABS, health data custodians) may want to monitor the FRIDGE project's published outputs for transferable governance approaches.
- Consider Policy teams working on secure compute access for government research could consider reviewing the full blog post and any associated technical documentation for relevant design patterns.
Risk, Assurance & Ethics2 items
Towards Risk-Aware Artificial Intelligence and Machine Learning Systems: An Overview
The MIT AI Risk Repository blog spotlights a 2022 academic paper by Zhang et al. that systematically organises AI and ML risks into two categories: data-level risks (including data bias, dataset shift, out-of-domain data, and adversarial attack) and model-level risks (including model bias, misspecification, and prediction uncertainty). The paper focuses on high-stakes decision settings and suggests drawing on reliability engineering concepts to develop risk-aware AI systems. It is the twenty-first framework catalogued in the MIT AI Risk Repository, which aims to consolidate diverse AI risk taxonomies into a single reference resource.
Key points
- A 2022 academic framework organises AI/ML risks into data-level and model-level categories with root causes and outcomes.
- The framework targets high-stakes decision settings like healthcare and transport - domains relevant to APS service delivery.
- This is a 2022 paper spotlighted by MIT's AI Risk Repository blog; it is not new primary research or Australian guidance.
Implications
- Monitor Risk and assurance teams may want to monitor the MIT AI Risk Repository as a consolidated reference for AI/ML risk taxonomies when developing or reviewing agency-level AI risk frameworks.
- Consider Agencies applying the APS Policy for the Responsible Use of AI could consider whether the data-level and model-level risk taxonomy maps usefully onto their existing AI risk assessment processes.
Why insurance companies should encourage solid AI risk management instead of excluding it
An OECD AI Wonk Blog post examines the emerging trend of major insurance providers seeking regulatory permission to exclude AI-related risks from coverage, citing companies such as AIG, Great American, and WR Berkley. The OECD's position is that insurers should instead use their market influence to encourage robust AI risk management practices. Only an excerpt is available in this item; the full argument and supporting evidence require direct engagement with the source. The piece is relevant context for APS practitioners thinking about how AI risk transfers and financial accountability frameworks evolve.
Key points
- Major US insurers including AIG are seeking to exclude AI-related risks from coverage, per OECD AI analysis.
- OECD argues insurers should instead incentivise sound AI risk management rather than exclude AI liabilities.
- Only a truncated extract is available; full argument and any policy recommendations are behind the link.
Implications
- Monitor Risk and procurement teams may want to monitor how AI exclusion clauses in insurance markets develop, as they could affect vendor liability and Commonwealth contract risk allocation.
- Consider Agencies developing AI risk frameworks could consider whether their vendor due diligence processes account for gaps in AI insurance coverage.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.