Item Catalogue
AI governance, regulation, strategy, and practice developments from monitored sources.
- Week of 23 September 2024
-
Evan Ratliff: Our Future with Voice Agents
- A journalist discusses his podcast experiment creating and deploying an AI voice clone of himself.
- Content focuses on journalism, personal experience, and societal musings rather than governance or policy.
- Low signal for APS readers; no governance, regulatory, or public-sector practice content.
- Week of 16 September 2024
-
Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements
- A 2023 survey catalogues seven core safety risks in large language models, from toxicity to malicious use.
- Covers evaluation techniques and improvement strategies across data preparation, training, and deployment phases.
- A research synthesis rather than policy guidance - useful as background reading, not directly actionable for APS.
- Week of 9 September 2024
-
Navigating the Landscape of AI Ethics and Responsibility
- A 2023 academic framework clusters AI ethics and responsibility issues into six groups via systematic literature review.
- The six clusters — broken systems, hallucinations, IP violations, privacy, malicious use, and environmental harm — map closely to APS AI risk categories.
- This is a blog summary of an existing academic paper; it is secondary source material with limited new content.
-
Meredith Ringel Morris: Generative AI's HCI Moment
- Google DeepMind's human-AI interaction research director discusses HCI, disability studies, and generative AI design.
- Topics include AI anthropomorphisation, AGI definitions, consent for generative 'ghost' systems, and bidirectional alignment.
- Academic podcast format with limited direct policy or governance takeaways for APS practitioners.
-
What's Missing From LLM Chatbots: A Sense of Purpose
- LLM benchmark saturation may not correlate with real-world user experience improvements in chatbot systems.
- Current evaluation methods are non-interactive and may poorly predict human-AI collaboration effectiveness.
- This is a short Substack preview with limited detail - the full argument requires reading the linked piece.
-
Statement by the Australian Public Service Commissioner on the Robodebt Centralised Code of Conduct Inquiry
- APSC found 12 individuals breached the APS Code of Conduct 97 times in connection with Robodebt.
- Robodebt involved algorithmic income-averaging automation, but this statement focuses on conduct accountability, not AI governance.
- Limited direct relevance to AI governance frameworks - this is a conduct and accountability item.
-
Mini-Update #47: First International AI Safety Treaty and the WavTokenizer Codec
- A paywalled newsletter covers the first international AI safety treaty and a technical audio codec development.
- The AI safety treaty thread may be relevant to APS policy teams tracking multilateral AI governance developments.
- Extracted text is too limited to assess substance - paywall prevents meaningful content analysis.
-
Update #83: AI Music Fraud and PlanSearch
- US federal prosecutors filed the first criminal charges for AI-assisted music streaming fraud, netting $10 million over seven years.
- AI scaled an existing fraud scheme but was incidental - the core crime was bot-driven royalty manipulation.
- PlanSearch, a new LLM code-generation algorithm from Scale AI researchers, is tangentially relevant to APS work.
- Week of 2 September 2024
-
Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems
- MIT AI Risk Repository catalogues a module-oriented LLM risk taxonomy covering 12 risks and 44 sub-categories.
- The framework covers input, model, toolchain, and output risks - a structure applicable to agency AI procurement and assurance.
- A research summary of an academic paper; not a government standard or mandatory framework.
-
Davidad Dalrymple: Towards Provably Safe AI
- ARIA Programme Director Davidad Dalrymple outlines a technical approach to provably safe, formally verified AI systems.
- ARIA's Safeguarded AI Programme is a direct UK government-funded research initiative on guaranteed AI safety - worth watching for Australian AISI.
- This is a podcast transcript with limited extractable detail; substantive engagement requires listening to the full episode.
- Week of 26 August 2024
-
TASRA: A Taxonomy and Analysis of Societal-Scale Risks from AI
- TASRA organises AI risks into six types based on accountability, intent, and scale of impact.
- The taxonomy's accountability-based framing aligns with APS responsible AI obligations around oversight and harm prevention.
- This is a 2023 academic preprint summarised in 2024 - useful for conceptual framing rather than immediate policy action.
-
Update #82: AI Lawsuits and SOPHON
- US courts are allowing copyright infringement claims against Midjourney, StabilityAI, and Anthropic to proceed.
- SOPHON is a research framework designed to prevent pre-trained AI models from being fine-tuned for harmful tasks.
- Item is a mixed newsletter combining legal commentary and technical research - moderate signal for APS readers.
- Week of 19 August 2024
-
Judy Fan: Reverse Engineering the Human Cognitive Toolkit
- A podcast interview with a Stanford cognitive scientist on human sensemaking and cognitive tools.
- AI is one methodological lens used in the lab's research, not the primary subject of discussion.
- Limited direct relevance to APS AI governance or policy work - included for completeness.
- Week of 12 August 2024
-
L.M. Sacasas: The Questions Concerning Technology
- A philosophy of technology podcast discussing human-technology relationships at a broad conceptual level.
- Covers outsourcing of skills, technological determinism, and what it means to be human - not AI-specific.
- Limited direct relevance to Australian federal agencies or AI governance practice.
- Week of 5 August 2024
-
AI Policy and Governance Newsletter — August 2024
- Australia's Criminal Code Amendment (Deepfake Sexual Material) Bill 2024 has passed the Senate and is now in force.
- US Senate bills on AI content protection, likeness rights, and financial regulatory sandboxes remain proposed and unvoted.
- Newsletter commentary is advocacy-leaning; APS readers should weigh analysis as think-tank perspective, not neutral assessment.
- Week of 29 July 2024
-
Lessons from the FDA for AI
- AI Now Institute draws on FDA pharmaceutical regulation to explore the case for ex ante AI oversight.
- The report maps FDA regulatory functions - premarket approval, post-market surveillance, enforcement - onto AI governance gaps.
- Published in mid-2024, it acknowledges that premarket AI enforcement remains a distant prospect in the US.
- Week of 8 July 2024
-
AI Policy and Governance Newsletter — July 2024
- Australian federal, state, and territorial governments jointly released an AI assurance framework now in force.
- Australia criminalised creation and sharing of non-consensual sexual deepfake images, with legislation before Parliament.
- Newsletter commentary notes the assurance framework does not meaningfully advance beyond 2022 ethical principles.
-
AI Now Co-ED Amba Kak Testifies at Senate Hearing on AI and Privacy
- AI Now Institute testimony argues US federal data privacy law is effectively AI regulation and essential now.
- Data minimisation, purpose limitation, and transparency rules are framed as structural AI governance tools.
- US-focused advocacy piece with limited direct APS applicability; useful as a comparative framing reference.
- Week of 1 July 2024
-
Public Interest AI for Europe? Shaping Europe’s Nascent Industrial Policy
- AI Now Institute frames European AI industrial policy as risking entrenchment of Big Tech rather than creating genuine competition.
- The piece raises transferable questions about public interest AI, market concentration, and sovereign digital infrastructure relevant to Australian policy.
- This is a research framing paper announcing future work - substantive findings are not yet published.
- Week of 24 June 2024
-
Safety and War: Safety and Security Assurance of Military AI Systems
- AI Now Institute argues military AI systems like Lavender and Gospel lack adequate safety assurance or independent oversight.
- The paper calls for safety engineering frameworks to evaluate military AI fitness-for-use against defined risk thresholds.
- Limited direct APS applicability now, but relevant to Defence AI governance and export control policy discussions.
- Week of 15 April 2024
-
AI Now co-ED Amba Kak’s Speech at the German Green Party’s Shaping AI Conference
- AI Now argues AI development is dominated by US and Chinese firms, with European AI startups unable to genuinely compete.
- Speech advocates for 'public interest AI industrial policy' as an alternative to replicating Big Tech models in Europe.
- This is a conference keynote with an advocacy frame - analytical substance is present but limited empirical grounding.
- Week of 11 March 2024
-
Power and Governance in the Age of AI
- A small number of Big Tech firms control AI infrastructure, research incentives, and de facto governance standards globally.
- Industry self-regulation fills the vacuum left by absent global coordination, raising accountability questions for national regulators.
- Published March 2024 by a US think tank - analytical framing, not policy with direct APS application.