Weekly AI Digest

1. DTA signs new 5-year agreement with Microsoft: Delivering value and innovation for the Australian Government

   AU 25 Feb 2026 Digital Transformation Agency

   The Digital Transformation Agency has secured a new five-year Volume Sourcing Agreement (VSA6) with Microsoft on behalf of the Australian Government, commencing 1 July 2026. The arrangement delivers stable pricing with capped increases, stronger legal protections for government data, and an explicit commitment to accelerating AI and emerging technology adoption across the APS. The agreement builds on findings from a recent review of Single Seller Arrangements, which found the existing arrangements had delivered $1.6 billion in discounts between 2019 and 2024. Smaller agencies are identified as particular beneficiaries of the standardised contracting framework.

   Implications

   • Monitor Agencies using Microsoft products could monitor DTA guidance on transition arrangements ahead of the 1 July 2026 commencement date.
   • Consider AI governance and procurement teams could assess how the agreement's enhanced data handling and legal provisions interact with existing agency-level AI risk and data governance frameworks.
   • Consider Agencies developing AI adoption roadmaps may want to consider how the VSA6's accelerated AI capability provisions affect their sourcing options and vendor dependency posture.

   Implications are AI-generated. Starting points, not advice.

   View details →

1. Deadline extension 20 March: Global call for ‘Governing with Artificial Intelligence’: Share your initiatives and insights on AI-driven innovation in government

   Global 1 Mar 2026 OECD AI Wonk Blog

   The OECD has launched a global call for contributions under the theme 'Governing with Artificial Intelligence', seeking government AI use cases, policy initiatives, and implementation tools to support trustworthy AI in public administration. The submission deadline has been extended to 20 March 2026. The extracted text is sparse and does not detail evaluation criteria, output format, or how contributions will be used. Australian agencies with mature AI governance or implementation examples may have an interest in participating.

   Implications

   • Consider Agencies with published AI use cases or governance frameworks could consider whether contributing to the OECD call aligns with their international engagement priorities.
   • Monitor Policy teams may want to monitor the resulting OECD collection as a source of comparative international government AI practice.

   Implications are AI-generated. Starting points, not advice.

   View details →

No primary items in this section.

No primary items in this section.

1. Generating Harms: Generative AI’s Impact and Paths Forward

   Global 24 Feb 2026 MIT AI Risk Repository – Blog

   The MIT AI Risk Repository has spotlighted the Electronic Privacy Information Center's 2023 paper 'Generating Harms', which presents a taxonomy of nine harm categories from generative AI: physical, economic, reputational, psychological, autonomy, discrimination, relationship, loss of opportunity, and dignitary. The paper draws on major AI harm taxonomies, provides real-world case studies covering deepfakes, misinformation, data breaches, and labour disputes, and discusses regulatory and industry interventions. The MIT blog post is a summary of an existing framework rather than new research, and the underlying paper reflects the policy landscape as of 2023.

   Implications

   • Consider APS risk and governance teams could assess whether EPIC's nine-category harm taxonomy maps usefully onto agency AI risk registers or existing harm assessment frameworks.
   • Monitor Teams tracking the MIT AI Risk Repository may want to monitor which of its 31+ catalogued frameworks are most cited in regulatory discussions, as this signals emerging international consensus on harm typologies.

   Implications are AI-generated. Starting points, not advice.

   View details →

1. Import AI 446: Nuclear LLMs; China's big AI benchmark; measurement and AI policy

   Global 23 Feb 2026 Import AI – Substack (Jack Clark)

   This edition of Import AI covers three distinct threads. First, Jacob Steinhardt argues that building technical measurement infrastructure is the single most tractable AI governance intervention, noting the field is talent-constrained and that measurement must precede effective policy. Second, a King's College London study finds LLMs used in simulated nuclear crises escalate more aggressively than humans, never choosing de-escalatory options and treating nuclear use as an ordinary strategic tool. Third, Chinese researchers have released ForesightSafety Bench, a comprehensive AI safety evaluation framework that closely parallels Western equivalents, suggesting convergence on safety evaluation norms despite geopolitical differences.

   Implications

   • Consider Agencies and AISI working on AI evaluation frameworks could consider Steinhardt's thesis that measurement investment is the highest-leverage governance intervention when prioritising capability uplift.
   • Monitor The nuclear wargame findings are worth monitoring as evidence of LLM behavioural risks in high-stakes advisory roles, relevant to any agency considering AI-assisted decision support.
   • Monitor ForesightSafety Bench's alignment with Western safety evaluation norms may be worth tracking as Australia engages in international AI safety cooperation.

   Implications are AI-generated. Starting points, not advice.

   View details →