Item Catalogue
AI governance, regulation, strategy, and practice developments from monitored sources.
- Week of 16 February 2026
-
Turning AI ambition into action: How the OECD is engaging at India’s AI Impact Summit
- OECD is participating in India's AI Impact Summit 2026, focusing on transparency and inclusive governance.
- OECD's engagement signals continued multilateral push to operationalise AI principles across emerging economies.
- Extracted text is minimal - substantive content unavailable; low signal for APS readers.
-
Using deep reinforcement learning to build better drift-aware malware detection
- Alan Turing Institute research applies deep reinforcement learning to malware detection that adapts as threats evolve.
- Drift-aware detection addresses a real operational gap where static models degrade as malware changes over time.
- Limited extracted content makes substantive assessment impossible - low signal for APS readers in current form.
-
Insights from our Digital Project Governance Boards launch session
- DTA released 'Steering for Success' guidance on digital project governance boards, developed with University of Queensland.
- Nine adaptable principles cover board composition, decision-making authority, risk alignment, and constructive culture.
- AI governance is not mentioned; this is general digital project governance guidance with limited direct AI relevance.
-
George Williamson appointed as CEO of the Alan Turing Institute
- Dr George Williamson CMG has been appointed as CEO of the Alan Turing Institute.
- The Alan Turing Institute is the UK's national institute for AI and data science.
- Leadership appointments rarely warrant priority attention for APS practitioners; low signal.
- Week of 9 February 2026
-
International Network for Advanced AI Measurement, Evaluation, and Science Publishes Consensus Areas on Practices for Automated Evaluations
- A ten-country network including Australia has published consensus practices for automated AI evaluation and measurement.
- Australia's participation signals alignment with emerging international standards on AI benchmarking and evaluation science.
- CAISI's draft Best Practices for Automated Benchmark Evaluations is open for public comment, offering a concrete reference point.
-
New study warns of risks in AI chatbots giving medical advice
- A randomised trial of 1,298 participants found LLMs no better than search engines for medical decision-making.
- Benchmark test performance does not reliably predict real-world safety - regulators and agencies should note this gap.
- Australian health agencies and AI governance teams considering LLM-assisted health tools have directly applicable evidence here.
-
New report calls for urgent action to tackle AI information threats following crisis events
- Alan Turing Institute report calls for urgent UK action on AI-generated misinformation following crisis events.
- Australian agencies managing public communications during emergencies face similar AI information threat risks.
- Extracted text is truncated - full report substance and specific recommendations are not available for assessment.
-
SafetyBench: Evaluating the Safety of Large Language Models
- SafetyBench is a bilingual benchmark evaluating LLM safety across 7 risk categories using 11,435 multiple-choice questions.
- The MIT AI Risk Repository spotlights this as one of 28 AI risk frameworks - useful for agencies mapping evaluation tools.
- The 2023 paper is research-vintage; the MIT blog post adds no new findings beyond summarising the original work.
-
Safety Assessment of Chinese Large Language Models
- A 2023 paper proposes a safety taxonomy for Chinese LLMs covering 8 harm scenarios and 6 adversarial attack types.
- The taxonomy is noted as scalable beyond Chinese-language models, with benchmarking across 15 LLMs including GPT series.
- This is a blog spotlight of an older paper via the MIT AI Risk Repository - limited immediacy for APS readers.
-
Import AI 444: LLM societies; Huawei makes kernels with AI; ChipBench
- Google-affiliated research finds LLM reasoning models simulate multiple internal personas, termed 'societies of thought.'
- ChipBench benchmark reveals frontier AI models perform poorly on real-world chip design tasks in Verilog.
- Both findings are foundational AI research with limited direct APS operational relevance at this stage.
-
The Global South can shape AI in practical terms: Why the India AI Impact Summit Matters
- OECD AI blog argues Global South nations can shape AI governance and real-world outcomes through forums like the India AI Impact Summit.
- Item is a short teaser post with minimal substantive content extracted - the full article may contain more detail.
- Limited direct relevance to APS readers; Australia is not a Global South actor in this framing.
-
Comment Now: Draft Guidelines on Data Classification Practices
- NIST SP 1800-39 provides practical guidance on classifying sensitive unstructured data using commercial tools.
- AI is mentioned only as a downstream beneficiary of good data classification - not the subject of the guidance.
- Limited direct relevance to Australian federal AI governance work; primarily a US data security standards item.
-
NIST Allocates Over $3 Million to Small Businesses Advancing AI, Biotechnology, Semiconductors, Quantum and More
- NIST allocated $3.19 million across eight small businesses under its SBIR Phase II program.
- AI features in only two of eight projects - one biopharmaceutical imaging tool and one cybersecurity compliance tool.
- Limited direct relevance to Australian federal AI governance, strategy, or policy work.
- Week of 2 February 2026
-
New Concept Paper on Identity and Authority of Software Agents
- NIST NCCoE is developing guidance on identity, authorisation, and access controls for agentic AI systems.
- The concept paper seeks public comment through April 2026 on use cases, standards, and challenges for AI agent IAM.
- Covers prompt injection mitigation and non-repudiation - emerging governance gaps directly relevant to APS AI deployments.
-
Model Evaluation for Extreme Risks
- A 2023 DeepMind-led paper proposes model evaluation frameworks targeting nine dangerous AI capability categories.
- The framework covers cyber-offense, deception, manipulation, weapons acquisition, and self-proliferation as extreme risk vectors.
- Primarily a research synthesis by MIT AI Risk Repository; the underlying paper predates recent Australian AI safety evaluation work.
-
The Ethics of Advanced AI Assistants
- Google DeepMind researchers systematically map ethical and societal risks of advanced AI assistants across three domains.
- The paper identifies an 'evaluation gap' where current assessments focus on models rather than broader sociotechnical systems.
- The framework is one of 24 catalogued in the MIT AI Risk Repository, useful as reference material rather than operational guidance.
-
AI Verify Testing Framework
- Singapore's AI Verify Testing Framework covers 11 ethical AI principles across transparency, safety, fairness, and oversight.
- The framework is aligned with ASEAN, EU, OECD, and US AI governance frameworks, giving it cross-jurisdictional relevance.
- This is a MIT blog spotlight of a 2023 Singapore government framework - useful context but not new guidance.
-
Import AI 443: Into the mist: Moltbook, agent ecologies, and the internet in transition
- Moltbook, an AI-agent-only social network, demonstrates emergent large-scale agent-to-agent interaction in the wild.
- A workshop report warns automated AI R&D could reduce human oversight and create rapid strategic surprise for governments.
- Both developments are early-stage and speculative but illustrate governance gaps that APS policy work may eventually need to address.
- Week of 26 January 2026
-
Towards Best Practices for Automated Benchmark Evaluations
- NIST CAISI has released draft NIST AI 800-2, covering best practices for automated benchmark evaluations of language models.
- The draft targets AI deployers, developers, and third-party evaluators - including procurement specialists using evaluation reports.
- Public comment closes 31 March 2026; Australian agencies or AISI could submit input to shape these emerging international standards.
-
AI assurance key to unlocking AI adoption in defence and driving UK economic growth
- Alan Turing Institute research argues a thriving AI assurance marketplace is essential to UK defence AI adoption and economic growth.
- UK-focused framing, but AI assurance ecosystem development is directly relevant to Australia's own nascent AI assurance market.
- Extracted text is truncated; full argument and evidence base cannot be fully assessed from available content.
-
Import AI 442: Winners and losers in the AI economy; math proof automation; and industrialization of cyber espionage
- AI systems using general foundation models can now solve advanced mathematics and assist original research.
- Independent testing shows frontier LLMs can generate zero-day exploits, signalling rapid AI-enabled cyber threat escalation.
- A Stanford economist argues AI will exceed the economic impact of electricity and semiconductors - with major risk implications.
-
Top British AI expertise to help spark renewal of public services and bolster national security
- The Alan Turing Institute is partnering with UK government to apply AI expertise to public services and national security.
- The initiative signals a growing pattern of national AI institutes being embedded in government delivery - relevant context for Australia's AISI.
- Extracted text is truncated; full substance of the announcement is not available for assessment.
-
Beyond the hype: Oxford & Berlin study uncovers four faces of ChatGPT’s early adopters
- Oxford/Berlin study identifies four ChatGPT user archetypes: Enthusiasts, Naïve Pragmatists, Cautious Adopters, and Reserved Explorers.
- Three of four archetypes expressed significant privacy concerns yet continued using AI tools - the 'privacy paradox' finding.
- Research draws on 2022-era early adopter data; generalisability to current APS AI adoption contexts is limited.
-
Can AI weather forecasting boost food security in the Global South?
- A UK initiative aims to deploy AI-driven weather forecasting to improve food security in sub-Saharan Africa.
- The project focuses on democratising access to AI weather prediction for smallholder agricultural resilience.
- Limited direct relevance to Australian federal agencies; primarily a development-sector AI application.
-
SUSHI@NIST: Rolling Next-Generation Secure Hardware into Standards
- NIST's SUSHI Workshop targets hardware security standards across the semiconductor development lifecycle.
- Workshop aims to initiate a 'Semiconductor Development Life Cycle Security Framework' informing national strategy.
- AI hardware security is a listed theme but not the primary focus; this is principally a semiconductor/hardware standards event.