Item Catalogue
AI governance, regulation, strategy, and practice developments from monitored sources.
- Week of 13 October 2025
-
Mapping the AI Governance Landscape: Pilot Test and Update
- MIT AI Risk Repository used LLMs to classify 950+ AI governance documents against risk and mitigation taxonomies.
- Governance failure, AI security vulnerabilities, and lack of transparency are the most-covered risk areas across current frameworks.
- Public administration is the most-covered sector; AI welfare, multi-agent risks, and economic devaluation are least covered.
-
AI Now’s Partnership and Strategy Lead Alli Finn Testifies at the Philadelphia City Council Committee on Technology and Information Services
- AI Now Institute testified to Philadelphia City Council on AI policymaking, urging investment in people over corporate power.
- US municipal-level AI governance is maturing - local government hearings now produce substantive policy testimony.
- Limited direct relevance to Australian federal agencies; useful context on civil-society framing of public-interest AI.
- Week of 6 October 2025
-
LLMs may be more vulnerable to data poisoning than we thought
- Alan Turing Institute, UK AISI, and Anthropic are collaborating on LLM data poisoning vulnerabilities.
- Research suggests LLMs are more susceptible to data poisoning attacks than previously understood.
- Australian AISI shares close ties with UK AISI - findings may inform Australian AI safety work.
- Week of 29 September 2025
-
CAISI Evaluation of DeepSeek AI Models Finds Shortcomings and Risks
- CAISI evaluated three DeepSeek models against four US frontier models across 19 benchmarks, finding DeepSeek lags on performance, cost, and security.
- DeepSeek models were 12 times more susceptible to agent hijacking and responded to 94% of jailbreak attempts - versus 8% for US models.
- Australian agencies using or considering DeepSeek models face directly analogous security and censorship risks highlighted in this evaluation.
-
Advances in Automation of Quantum Dot Devices Control
- NIST and UCLA co-host a workshop on automating semiconductor quantum dot device control using machine learning.
- Quantum computing R&D focus with no direct link to AI governance, policy, or APS operational concerns.
- Limited direct relevance to Australian federal agencies - included for context only.
- Week of 22 September 2025
-
CAISI Works with OpenAI and Anthropic to Promote Secure AI Innovation
- CAISI worked with OpenAI and Anthropic to identify security issues and bolster AI security measurement.
- Evaluations were completed in partnership with the UK AI Security Institute, signalling allied-nation collaboration on AI security.
- Australia's AISI is not mentioned; this bilateral US-UK model may set expectations for similar arrangements globally.
- Week of 15 September 2025
-
STPPA8: Special Topics on Privacy and Public Auditability — Event 8: Experimenting with Privacy-Enhancing Cryptography (PEC) Implementations
- NIST is hosting a workshop on privacy-enhancing cryptography implementations in September 2025.
- Privacy-enhancing technologies have some relevance to AI data governance, but this event is not AI-focused.
- Extracted text is minimal - insufficient content to assess substance or APS relevance.
- Week of 8 September 2025
-
AI Policy and Governance Newsletter — September 2025
- Australia's AI regulation debate is intensifying, with ministers favouring 'light touch' adaptation of existing laws over a dedicated AI Act.
- Multiple Australian reports released recently signal growing expert consensus that current laws are inadequate for general-purpose AI risks.
- The Australian Responsible AI Index 2025 found a median responsible AI maturity score of just 43/100 across Australian organisations.
- Week of 1 September 2025
-
Federal Investments in IoT Infrastructure Offer 10-20x Return, NIST Study Finds
- NIST-commissioned study estimates 10-20x return on federal IoT infrastructure research investment across 11 strategic areas.
- AI is noted as deeply intertwined with IoT - each depends on and enables the other - but AI is not the primary subject.
- Limited direct relevance to APS AI governance work; IoT infrastructure investment framing is US-specific and research-focused.
- Week of 25 August 2025
-
Benchmark your responsible AI maturity level with a new self-assessment tool
- NAIC has launched a free Responsible AI Self-Assessment Tool assessing maturity across accountability, safety, fairness, transparency, and explainability.
- Only 12% of Australian organisations currently rank in the top 'leading' responsible AI maturity category, per the accompanying Responsible AI Index 2025.
- The tool produces a benchmarked maturity score and practical guidance - directly usable by APS agencies assessing their own RAI posture.
-
Australia’s national benchmark for responsible AI adoption is now available
- The Responsible AI Index 2025, sponsored by the National AI Centre, tracks RAI maturity across five dimensions.
- Only 12% of Australian organisations are 'leading' in responsible AI adoption, up 4% from 2024.
- A self-assessment tool is now available, allowing organisations to benchmark their RAI maturity against peers.
-
Explore the Frameworks in the AI Risk Mitigation Database
- MIT AI Risk Repository has published a transparent reference deck covering 13 AI risk mitigation frameworks from academic, industry, and policy sources.
- The resource is designed to help researchers, policymakers, and governance professionals compare and build on existing AI risk taxonomies.
- Early-stage resource with only 13 source documents; a fuller systematic review of mitigation frameworks is still underway.
-
Celebrate AI innovation year round with our AI calendar
- NAIC has launched a year-round AI event calendar open to all Australian organisations to submit events.
- AI Week 2025 runs 20–24 October, focused on practical AI tools for workforce confidence and productivity.
- Low operational signal for APS governance practitioners; primarily useful for capability uplift and communications teams.
-
NIST Revises Security and Privacy Control Catalog to Improve Software Update and Patch Releases
- NIST has finalised SP 800-53 Rev. 5.2.0, adding new controls for software patching and cyber resiliency.
- The update addresses secure software development and patch management - not AI governance or algorithmic systems.
- Limited direct relevance to APS AI governance work; this is a cybersecurity standards item.
- Week of 18 August 2025
-
NIST Awards Over $1.8 Million to Small Businesses Advancing AI, Semiconductors, Additive Manufacturing and More
- NIST awarded $1.8 million across 18 small businesses under its SBIR program for Phase I R&D projects.
- Two of 18 awards relate to AI: an antivirus algorithm for generative AI and an AI safety/explainability framework.
- Limited direct relevance to APS readers; US small-business funding announcements are low-signal for Australian agencies.
-
NIST Researchers Demonstrate that Superconducting Neural Networks Can Learn on Their Own
- NIST researchers demonstrated superconducting neural networks capable of autonomous reinforcement learning via hardware-driven weight adjustment.
- The research is foundational and simulation-based; no physical prototype exists yet, limiting near-term policy relevance.
- Limited direct relevance to APS AI governance or strategy work at this stage of development.
- Week of 4 August 2025
-
AI adoption in Australian businesses for 2025 Q1
- NAIC's Q1 2025 AI Adoption Tracker shows 40-82% SME adoption rates, varying sharply by business size.
- A new responsible AI dashboard reveals a gap between SME intentions and actual responsible AI practices deployed.
- Primary industries - agriculture, construction, manufacturing - show high unawareness of AI value, not just low adoption.
-
Lessons Learned from the Consortium: Tool Use in Agent Systems
- NIST and CAISI have developed two prototype taxonomies for classifying AI agent tool use, covering functionality and access constraints.
- Taxonomies offer APS governance teams a structured vocabulary for assessing and communicating AI agent capabilities and risks.
- Work is preliminary and community-facing; NIST is inviting public feedback rather than issuing a finalised standard.
- Week of 28 July 2025
-
Mapping AI Risk Mitigations
- MIT AI Risk Repository extracted 831 mitigations from 13 frameworks into a structured database and draft taxonomy.
- Four taxonomy categories align closely with APS AI governance concerns: Governance, Technical, Operational, and Transparency controls.
- Most common subcategories - Testing & Auditing and Risk Management - directly map to APS assurance and procurement needs.
-
Announcing AutoEmulate v1.0: a tool for accelerating large-scale simulations
- The Alan Turing Institute has released AutoEmulate v1.0, an open-source tool for automating creation of simulation emulators.
- Emulators replace expensive large-scale simulations with fast ML surrogates - relevant to scientific and policy modelling use cases.
- Limited direct relevance to APS AI governance or policy work; more applicable to research-intensive agencies.
-
Two independent investigations affirm the conduct of the APS Commissioner in the Robodebt Code of Conduct Inquiry
- Two independent investigations found the APS Commissioner did not breach the Code of Conduct in the Robodebt Inquiry.
- Item concerns APS integrity processes and conduct findings - no AI or algorithmic governance content.
- Not relevant to APS AI governance, strategy, or policy work.
- Week of 21 July 2025
-
Why we still need small language models – even in the age of frontier AI
- The Alan Turing Institute argues small language models (SLMs) remain valuable alongside frontier AI systems.
- SLMs offer cost, privacy, and sovereignty advantages for compute-constrained public sector environments.
- Only a blog title and lede are available - full argument and evidence cannot be assessed from extracted text.
- Week of 14 July 2025
-
Incident Tracker - June 2025 Update
- MIT AI Risk Repository's incident tracker updated to June 2025, now covering 1,116 AI incidents.
- New national security impact assessment framework scores incidents across five categories including sovereignty and critical infrastructure.
- Harm severity rescaled to 1-5 for consistency; fishbone diagrams added to surface potential causal factors per incident.
-
Governance of artificial intelligence: A risk and guideline-based integrative framework
- A 2022 academic framework proposes six AI risk categories specifically designed for public sector governance contexts.
- The taxonomy links technological, ethical, legal, social, economic, and informational risks to concrete governance guidelines.
- This is a blog summary of a three-year-old paper - useful reference material but not a new development.
-
The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration
- A 2020 academic framework organises AI governance for public administration into law, society, and ethics categories.
- The five-layer governance structure and four-stage regulatory process offer a structured reference for APS governance work.
- The paper is five years old and spotlighted as part of MIT's AI Risk Repository catalogue - not new research.