Australia establishes new institute to strengthen AI safety
25 November 2025
The Australian Government will establish the Australian Artificial Intelligence Safety Institute (AISI) to address AI-related risks and harms. The goal is simple: provide a trusted, expert capability inside government to test, monitor and share insights on advanced AI systems.
AISI will help government spot future risks early and ensure protections keep pace with the tech. It becomes operational in early 2026.
What AISI will do
- Track fast-moving AI developments and respond to emerging risks and harms.
- Deepen government's technical grasp of advanced AI and its potential impacts.
- Operate as a central hub to coordinate insights and action across agencies.
- Provide guidance on AI opportunity, risk and safety to business, government and the public via established channels, including the National AI Centre (NAIC).
- Support Australia's commitments under international AI safety agreements and join the International Network of AI Safety Institutes (INASI).
Why this matters for government teams
AISI gives policy, risk, procurement and digital teams a technical backbone for AI assurance. Expect clearer guidance on testing practices, model evaluations, and how to handle high-risk use cases across public services.
It will complement existing legal and regulatory settings that protect rights and safety. Think of it as added capability, not a replacement for current obligations.
What agencies can do now
- Map AI use cases in flight and flag those with potential safety, security or privacy risks.
- Identify datasets, models and third-party services that may require assurance or independent testing.
- Align procurement templates with AI risk controls (evaluation criteria, red-teaming, incident reporting, and kill-switch requirements).
- Stand up a point of contact to engage with AISI and the NAIC once guidance and testing services roll out.
- Refresh internal policies on model usage, human oversight, and data handling for generative features.
How AISI fits with existing frameworks
The institute's work will sit alongside privacy, consumer protection and online safety obligations already in force. Agencies should continue applying current risk, security and ethics processes, with AISI adding technical testing and shared intelligence across government.
Timeline
AISI becomes operational in early 2026. Initial priorities are expected to focus on testing methods, cross-agency coordination, and guidance for higher-risk deployments.
Get ready and upskill
If your team is planning pilots or procurement involving advanced AI, invest in practical skills now. For role-based learning paths and certifications, see Complete AI Training: Courses by Job.
Your membership also unlocks:
