Australian AI Safety Institute

The Australian AI Safety Institute (AISI) was announced on 25 November 2025 by the Australian Government as part of its National AI Plan, with operations commencing in early 2026. It sits within the Department of Industry, Science and Resources as a whole-of-government hub for AI safety, providing trusted expert capability to monitor, test, and share information on emerging AI technologies, risks, and harms.

The AISI operates through three primary work streams. First, it conducts pre-deployment testing of frontier AI models through voluntary evaluations using methodologies such as red teaming and safety case review. Second, it performs ongoing risk monitoring, including both upstream assessment at the design stage and downstream analysis of real-world impacts. Third, it engages in information sharing with government agencies, industry stakeholders, and international partners.

The institute assesses a range of AI risks including CBRN misuse, enhanced cyber capabilities, loss-of-control scenarios, information integrity and influence risks, and broader systemic risks from general-purpose AI systems. Its insights support ministers and regulators in maintaining fit-for-purpose safety measures and legal frameworks that keep pace with rapid technological change.

Critically, the AISI is not a regulator and cannot compel compliance. Specialist regulators retain enforcement authority under existing Australian law. The AISI's role is to provide technical expertise and recommendations that enable government bodies to act.

Australia is a founding member of the International Network for Advanced AI Measurement, Evaluation and Science. Anthropic has also signed a Memorandum of Understanding with the Australian Government to cooperate on AI safety research and to work directly with the AISI, including sharing findings on model capabilities and risks and participating in joint safety evaluations.

The AISI was funded with AUD $29.8 million over four years from 2025-26, with AUD $7.9 million per year ongoing from 2029-30. The founding team was recruited across senior leadership, AI safety research scientists, AI safety engineers, and AI risk specialists, with positions open Australia-wide under flexible arrangements.