Public research body · Science policy · Canberra, Australian Capital Territory (Industry House
The Australia AI Safety Institute (AISI) is an Australian Government capability within the Department of Industry, Science and Resources. Its stated mandate is to analyze and test emerging frontier AI capabilities and assess associated risks and harms, while providing technical information to support regulators and agencies and helping shape Australia’s approach to safe AI development, deployment and international governance.
Functionally, AISI focuses on (1) monitoring and analyzing frontier models and new system types such as AI agents and highly capable generative models, (2) monitoring real-world harms and emerging risks from AI use in society and the economy, and (3) contributing to international scientific work on AI safety measurement, evaluation and governance, including through the International Network for Advanced AI Measurement, Evaluation and Science (NAAIMES) and the International AI Safety Report ecosystem. AISI also emphasizes operational support to other government bodies: it aims to provide technical insights that help regulators keep pace with AI developments, and it coordinates with domestic entities such as Australia’s National AI Centre (NAIC) to translate AI Safety Institute insights into practical guidance for businesses, organizations and the public.
No indexed openings right now. Check the careers page ↗.
Publications are not yet indexed for this organization. Read them on industry.gov.au ↗.
AISI commissioned Gradient Institute to produce an analytical framework addressing risks and controls when AI agents interact across organizational boundaries, including identifying deployment tiers and gaps in governance/control responsibilities.
Australia and UK strengthen partnership on safe and secure AIAustralia and the UK announced a memorandum of understanding (MoU) to strengthen technical cooperation between the Australian AI Safety Institute (currently being established) and the UK AI Security Institute, including shared information/expertise, joint research approaches to measuring/testing/managing AI risks, and support for multilateral work of NAAIMES.
Australia meets with global partners on AI evaluation and measurementAISI covered participation in the International Network for Advanced AI Measurement, Evaluation and Science (NAAIMES) meeting to share knowledge and align on best practice for AI evaluation and measurement, including workshops on best principles and approaches for conducting AI evaluations.