Week of 5 January 2026
Meta's KernelEvolve uses LLMs to automate AI kernel design, cutting development time from weeks to hours.
Key points
- Epoch AI analysis finds decentralised AI training growing at 20x per year, raising governance and sovereignty implications.
- Item is a technical research newsletter; policy implications are present but require significant extrapolation for APS use.
Alan Turing Institute blog advocates using AI and data science to address sustainability challenges.
Key points
- Extracted text is too sparse to assess specific claims, methods, or findings in detail.
- Limited direct relevance to Australian federal AI governance - primarily a UK think-tank perspective piece.
Week of 22 December 2025
Stanford/CMU research shows AI agents with scaffolding match professional penetration testers at $18/hour versus $60/hour for humans.
Key points
- The ARTEMIS framework demonstrates frontier AI systems are systematically under-elicited - more capable than they appear without structured scaffolding.
- Remaining items cover robotics data transfer (OSMO glove) and AI-assisted chip design - limited direct APS relevance.
MLCommons AI Safety Benchmark v0.5 defines 13 hazard categories for evaluating chat-based AI system safety.
Key points
- Practical testing tools including ModelBench are openly available, making this usable for agency-level AI evaluation.
- V0.5 has been superseded by V1.0 (AILuminate, Feb 2025); this spotlight is retrospective context, not a new release.
MIT AI Risk Repository spotlights a 2023 paper categorising catastrophic AI risks into four proximate causes.
Key points
- The four categories — malicious use, AI race, organisational risks, and rogue AI — each include mitigations.
- This is a secondary blog summary of a 2023 paper; primary value is as a reference for risk taxonomy work.
Week of 15 December 2025
NIST has released a preliminary draft Cyber AI Profile (NISTIR 8596) for 45-day public comment, closing 30 January 2026.
Key points
- 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.
NIST has released a preliminary draft Cybersecurity Framework Profile for AI (NIST IR 8596) open for public comment until 30 January 2026.
Key points
- The profile addresses three focus areas: securing AI system components, AI-enabled cyber defence, and thwarting AI-enabled attacks.
- A companion workshop is scheduled for 14 January 2026; this is a US standard with no immediate Australian compliance obligation.
A 2022 academic framework organises AI/ML risks into data-level and model-level categories with root causes and outcomes.
Key points
- 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.
Major US insurers including AIG are seeking to exclude AI-related risks from coverage, per OECD AI analysis.
Key points
- 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.
The Alan Turing Institute's FRIDGE project enables secure research using sensitive data on AI supercomputers.
Key points
- 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.
University of Birmingham and Alan Turing Institute won £610,000 to develop AI space-threat detection tools.
Key points
- The project targets national security applications - a domain of growing interest to Australian defence and intelligence agencies.
- Limited direct relevance to APS AI governance practitioners; more pertinent to defence science and space policy communities.
NIST CAISI is hiring an AI Research Scientist focused on evaluation methods and trustworthy AI measurement.
Key points
- The role signals continued US investment in rigorous AI evaluation infrastructure relevant to international standards work.
- Job postings carry low signal for APS readers; included for context on CAISI's capability build-out only.
NIST has finalised cybersecurity guidelines for smart speaker use in home telehealth and hospital-at-home programs.
Key points
- Guidelines draw on NIST CSF 2.0, Privacy Framework, and IoT baseline standards - not AI governance frameworks specifically.
- Limited direct relevance to APS AI governance work; this is primarily a health-sector IoT/cybersecurity item.
Week of 8 December 2025
Good Ancestors' December 2025 newsletter covers Australia's National AI Plan, AISI announcement, and multiple international developments.
Key points
- The Australian AISI ($30m, early 2026 start) and National AI Plan are the headline items, with APS AI Plan also featured.
- Newsletter spans AI espionage, model releases, deepfake legislation, CSIRO restructure, and EU AI Act delays across 15+ distinct items.
Import AI's issue 437 covers four distinct topics: co-improving AI, AI labelling policy complexity, SimWorld simulator, and DeepMind's SIMA 2 agent.
Key points
- The AI labelling section directly illustrates why simple-sounding AI policy can impose significant compliance burdens on industry.
- Coverage is research-forward and internationally focused; limited direct APS operational relevance but useful as a frontier signal.
Week of 1 December 2025
DTA has released an updated AI policy, a new AI Impact Assessment Tool, and new AI procurement guidance, effective 15 December 2025.
Key points
- The updated Policy mandates AI impact assessments for all use cases and requires agencies to develop and communicate a strategic position on AI adoption.
- An AI Review Committee for high-risk use cases across the APS is being finalised, with terms of reference expected in Q1 2026.
The National AI Centre released practical guidance on labelling, watermarking, and metadata for AI-generated content.
Key points
- Guidance targets businesses but applies equally to APS agencies producing AI-assisted communications and official documents.
- Framed around regulatory risk reduction and trust-building, aligned with responsible AI use principles in government.
MIT AI Risk Repository Version 4 now includes over 1,700 coded risks drawn from 74 published frameworks.
Key points
- Nine newly added frameworks span government reports, peer-reviewed papers, and industry sources, including a UK DSIT frontier AI paper.
- A structured, living reference for AI risk taxonomy - useful for APS governance and risk assessment work.
Alan Turing Institute research identifies new AI risks specific to financial sector institutions.
Key points
- Financial sector AI risk findings are relevant to Australian agencies managing financial data or payment systems.
- Extracted text is truncated - full research scope and findings are not assessable from available content.
Alan Turing Institute researchers have developed an AI model enabling real-time maritime surveillance onboard satellites.
Key points
- Onboard processing removes the need to downlink raw imagery, reducing latency and bandwidth demands significantly.
- Limited direct relevance to APS governance practice; primarily a technical research item from a UK institution.
Week of 24 November 2025
Import AI #436 covers four distinct topics: AI agent training infrastructure, a 2GW datacenter in Saudi Arabia, regulation critique, and a RAND paper on countering rogue AI.
Key points
- The RAND analysis on countering a rogue superintelligence offers sobering conclusions relevant to AI safety policy discussions.
- The regulation critique reflects a pro-innovation perspective on regulatory burden - a counterpoint worth noting in AI governance debates.
Google DeepMind philosopher Iason Gabriel discusses value alignment, distributive justice, and ethics of advanced AI assistants.
Key points
- Topics include aligning LLMs with democratic norms, AGI social power dynamics, and the challenge of AI value alignment at scale.
- A podcast interview format - conceptually rich but not directly actionable for APS practitioners without further engagement.
Week of 17 November 2025
The APS AI Plan has launched, jointly led by Finance, APSC, and DTA, structured around Trust, People, and Tools pillars.
Key points
- Agencies must appoint Chief AI Officers, designate accountable officers per use case, maintain internal AI registers, and conduct AI impact assessments.
- A new AI Review Committee managed by DTA will provide cross-government scrutiny of high-risk AI use cases.
Week of 10 November 2025
AI industry energy demands are driving pressure to fast-track nuclear deployment, undermining established safety regulation.
Key points
- LLMs being proposed for nuclear licensing documents raise proliferation and cybersecurity risks with unsubstantiated efficiency claims.
- Limited direct APS relevance; Australia lacks operational nuclear power, though SMR policy interest is growing.
Week of 3 November 2025
Good Ancestors' October/November 2025 newsletter covers a dense cluster of Australian and international AI policy developments.
Key points
- Key Australian threads include Treasury's light-touch ACL review, eSafety's new AI chatbot codes, and the pending National AI Plan.
- Australia's IMD digital competitiveness ranking fell from 15th to 23rd, with its AI laws ranking dropping from 8th to 34th in one year.