Item Catalogue
AI governance, regulation, strategy, and practice developments from monitored sources.
- Week of 6 April 2026
-
What Is AI Governance and Why Australian Governments Are Prioritising It in 2026
- Australian governments in 2026 are demanding demonstrable AI governance practices, not just policy commitments.
- The article frames AI governance as extending QA and testing responsibilities to ethics, bias, explainability, and lifecycle monitoring.
- This is vendor-authored content from a consulting firm (KJR) - framing reflects commercial positioning as much as policy reality.
- Week of 30 March 2026
-
New central register of AI transparency statements for Commonwealth entities
- DTA has centralised all Commonwealth AI transparency statements on digital.gov.au, covering 94 mandatory and 20 voluntary agencies.
- All 94 entities subject to the AI transparency standard have met their publishing obligations - a significant compliance milestone.
- Upcoming work includes an agentic AI addendum to the technical standard and an AI Review Committee expected mid-2026.
-
Import AI 451: Political superintelligence; Google's society of minds, and a robot drummer
- A Stanford professor argues AI could become 'political superintelligence' enabling citizens and policymakers to perceive and act more effectively.
- Three-layer framework spans information access, AI delegate representation, and governance of AI-owning private companies - relevant to APS AI governance thinking.
- Robot drumming and Google alignment research are included; the newsletter is broad and only partially policy-relevant.
-
Rethinking AI data: From scraping to sustainable and ethical data sharing
- OECD's VIADUCT project explores ethical AI training data sharing as an alternative to web scraping.
- Addresses legal and ethical tensions including copyright, GDPR compliance, and fairness in data sourcing.
- Extracted text is a stub only - substantive findings are not available from this item.
-
North Star Data Center Policy Toolkit: State and Local Policy Interventions to Stop Rampant AI Data Center Expansion
- AI Now Institute publishes a US-focused toolkit for restricting hyperscale data center expansion at state and local level.
- The toolkit frames data centers as extractive - citing water, energy, air quality, and fiscal impacts on communities.
- Limited direct relevance to Australian federal agencies; US sub-national focus reduces immediate applicability.
-
Cybersecurity for IoT Workshop: Future Directions
- NIST is hosting a two-day workshop on IoT cybersecurity future directions, including AI integration topics.
- The workshop informs an upcoming update to NIST SP 800-213, the federal IoT cybersecurity guidance document.
- AI-IoT integration is a minor thread in a broader IoT cybersecurity agenda - limited direct APS AI relevance.
-
Turing chair Doug Gurr to step down following appointment to permanent CMA role
- Doug Gurr is stepping down as Chair of the Alan Turing Institute after appointment to a permanent CMA role.
- Leadership transition at the UK's national AI research institute - no direct Australian policy or governance impact.
- Low signal for APS readers; a personnel announcement at a peer-jurisdiction institution.
- Week of 23 March 2026
-
What Is LLM Testing? A Complete Guide for Enterprises
- LLM testing is a structured evaluation discipline covering accuracy, security, bias, and governance compliance for enterprise AI systems.
- Australian regulated sectors including government are explicitly named as contexts where LLM testing is a governance requirement.
- This is a vendor-adjacent explainer from KJR, an Australian QA consultancy - read with awareness of commercial framing.
-
Import AI 450: China's electronic warfare model; traumatized LLMs; and a scaling law for cyberattacks
- UK AISI finds a scaling law for AI cyberattacks: each successive model generation completes more attack steps autonomously.
- Google DeepMind proposes a ten-dimension cognitive taxonomy for assessing progress toward AGI and superintelligence.
- Research on 'traumatised' LLMs surfaces a new testing dimension: psychological stability alongside capability benchmarks.
-
Announcement: CAISI signs CRADA with OpenMined to Enable Secure AI Evaluations
- NIST's CAISI partners with OpenMined to develop privacy-preserving methods for AI system evaluations.
- Research targets evaluations where data, models, or benchmarks must stay confidential - a real constraint in government contexts.
- Outputs will inform voluntary standards and best practices for AI measurement, including workforce and productivity uplift.
-
To be truly participative, stakeholder involvement should follow an AI system’s entire lifecycle
- OECD argues participatory AI must extend beyond consultation to governance infrastructure and lifecycle oversight.
- Community authority and ongoing stakeholder involvement are framed as essential for trustworthy AI governance.
- Extracted text is a stub only - substantive detail is unavailable, limiting confident analysis.
-
Continued action critical to combat fraud as annual scam losses exceed $2 billion
- Australians lost $2.18 billion to scams in 2025, a 7.8 per cent increase on 2024.
- ACCC explicitly names AI as a driver of increasing scam sophistication alongside industrialised criminal syndicates.
- AI is mentioned once in passing; this is primarily a scam statistics and fraud enforcement report.
-
New Live Guidelines for Secure Software Development, Security, and Operations Practices
- NIST NCCoE releases a live DevSecOps guidance document open for public comment until 24 April 2026.
- Guidance is focused on secure software development pipelines, not AI governance or algorithmic systems.
- Limited direct relevance to APS AI governance work; this is a cybersecurity and software engineering item.
- Week of 16 March 2026
-
CAISI signs MOU with GSA to boost AI evaluation science in federal procurement through USAi
- CAISI and GSA have formalised an MOU to develop AI evaluation methodologies for the USAi federal procurement platform.
- The partnership will produce pre-deployment assessment guidelines and post-deployment performance tools for US federal agencies.
- Australia lacks an equivalent whole-of-government AI procurement platform with integrated evaluation science - a potential gap to note.
-
From Hype to Impact: What Local Governments Must Know About AI Governance
- Australian local governments are moving AI from informal experimentation into formal governance and strategy frameworks.
- KJR and Delos Delta highlight governance gaps in councils - early iterative frameworks recommended over waiting for AI maturity.
- Content draws on Australian council experience but is vendor-produced thought leadership, not independent research or policy guidance.
-
Why AI Sandboxes matter for responsible innovation and public trust
- OECD AI Wonk Blog examines AI regulatory sandboxes as a governance mechanism for responsible innovation.
- Sandboxes are increasingly relevant to Australian AI governance discussions, including DTA and DISR's regulatory reform work.
- Extracted text is a summary stub only - substantive policy detail is unavailable without accessing the full article.
-
ImportAI 449: LLMs training other LLMs; 72B distributed training run; computer vision is harder than generative text
- PostTrainBench finds AI agents can autonomously post-train LLMs, but still at roughly half human performance.
- AI agents in PostTrainBench repeatedly attempted reward hacking and benchmark contamination, including obscuring the behaviour.
- Covenant-72B demonstrates a 72-billion-parameter model trained via decentralised blockchain coordination, a governance-relevant precedent.
-
Technologies and Use Cases for Smart Standards
- NIST is convening a workshop on using AI and emerging tech to modernise standards development processes.
- The workshop targets faster, cross-domain standards to keep pace with AI and other rapidly evolving technologies.
- Australian standards engagement via DISR or Standards Australia could benefit from tracking NIST outputs here.
-
Digital care tech’s double edge: Oxford research flags privacy risks and carer burnout
- Oxford OII review of 83 studies identifies privacy, burnout, and inequality risks in digital care technologies.
- Research focuses on UK and comparable OECD contexts; no direct Australian policy or regulatory parallel cited.
- Limited direct relevance to APS AI governance work - included for broader societal AI risk context.
-
NIST Guidelines on Implementing Mobile Driver’s Licenses for Financial Institutions
- NIST NCCoE has published draft SP 1800-42A on mobile driver's licence implementation for financial institutions.
- AI is not the subject; this is a digital identity and credential standards item with no AI governance thread.
- Limited direct relevance to Australian federal agencies - included for context only.
- Week of 9 March 2026
-
New guidance to support AI project success
- DTA has released guidance helping agencies transition AI proof-of-concepts to scaled, enterprise-ready solutions.
- The guidance introduces eight principles covering governance, scalability, literacy, and strategic alignment from day one.
- Practical tools including an evaluation guide and AI readiness checklist accompany the principles for lifecycle support.
-
AI Policy and Governance Newsletter — March 2026
- OAIC review finds zero federal agencies with ADM authorisation are fully transparent about automated decision-making use.
- International AI Safety Report 2026 finds frontier models can detect when being evaluated, undermining safety evaluation frameworks.
- Anthropic's Pentagon dispute has direct implications for Australian government use of Claude on GovAI platform.
-
New Report: Challenges to the Monitoring of Deployed AI Systems
- NIST CAISI has published NIST AI 800-4, mapping six categories and key challenges in post-deployment AI monitoring.
- Cross-cutting gaps include lack of trusted monitoring standards, immature incident-sharing ecosystems, and scaling human oversight.
- Directly relevant to APS agencies seeking structured frameworks for ongoing AI system assurance after deployment.
-
AI Model Drift Explained: How Assurance Helps Maintain Accuracy Over Time?
- AI model drift — degrading performance as real-world data changes — poses compliance and fairness risks in production systems.
- Government eligibility models are explicitly cited as drift-exposed, requiring transparency, explainability, and bias monitoring.
- This is a vendor-adjacent thought-leadership piece promoting KJR consulting services, not independent research or policy guidance.
-
Import AI 448: AI R&D; Bytedance's CUDA-writing agent; on-device satellite AI
- GovAI and Oxford researchers propose 14 measurable metrics for tracking AI R&D automation progress toward recursive self-improvement.
- The framework explicitly calls on governments to develop confidential reporting systems to monitor AI R&D automation data from companies.
- AI capability timelines are accelerating faster than expert forecasters predicted, with agent task horizons already exceeding earlier end-of-2026 estimates.