Preparing for AGI¶

Artificial general intelligence (AGI) refers to AI systems that can match or exceed human-level performance across a wide range of cognitive tasks, rather than excelling at just one. Australia may deploy increasingly capable systems across defence, critical infrastructure, finance, health and media without building or fully verifying those systems ourselves.

Most AI guidance focuses on near-term adoption, governance and compliance. This section complements that work with possible futures, Australian risk scenarios and the C·A·G·R framework for structured preparation across uncertain timelines.

Time to explore: 30–45 minutes | Audience: Government, business leaders, community organisations

Why does Australia need AGI preparedness?¶

What happens when Australian organisations depend on systems developed overseas, cannot fully inspect them and still remain accountable for the consequences? That question matters before any particular definition of AGI is reached.

Australia is a small, advanced economy that imports most frontier AI while deploying it across finance, health, government and critical infrastructure. This creates a distinctive mix of exposure and influence. (What is frontier AI?)

Important features of the Australian context:

• Geographic distance and long supply chains can increase exposure to external disruption, while also creating incentives for domestic continuity planning
• Concentrated sectors, including banking, media and energy, can turn a provider or system failure into a broader sectoral problem
• Five Eyes and regional partnerships give Australia channels for technical cooperation, intelligence sharing and coordinated policy
• Established regulators and public institutions provide foundations for accountability, although capability and authority vary by sector
• Universities and research organisations can contribute to independent evaluation and safety research
• Procurement, sectoral regulation and deployment decisions give Australia practical influence over systems used here, even when their development happens elsewhere

This section addresses Australia's specific context—not the US, UK or EU.

What strategic challenges does Australia face?¶

Australia is likely to deploy advanced AI systems that were developed overseas and cannot be fully verified by local users or regulators. This creates three connected strategic problems:

The information problem: We can't verify what we import¶

Frontier AI providers assure us their systems are safe and aligned. But we face fundamental information asymmetries: we don't observe training processes, current techniques cannot reliably determine what systems are optimising for and systems may appear aligned in testing but pursue different goals in deployment (deceptive alignment). Even well-intentioned providers cannot provide strong, independently verifiable guarantees of safety today—alignment verification remains an open technical challenge.

The coordination problem: International cooperation is hard¶

Many protective measures work best with international cooperation, including compute governance, export controls and safety standards. Coordination nevertheless faces structural barriers: competitive dynamics can create pressure to deploy before safety validation, nations with frontier AI industries may have different incentives from importers and treaty processes can move slowly. Australia can contribute through Five Eyes partnerships and multilateral forums, but domestic plans should not depend on timely international agreement.

Time as a constraint: Capability and governance move at different speeds¶

Some AI capabilities and deployment practices are changing faster than legislation, institutional expertise and assurance methods. The practical concern is not only when AGI might arrive, but whether evaluation, governance and continuity arrangements can keep pace with the systems organisations choose to deploy.

These problems justify early, proportionate preparation. The C·A·G·R Framework uses defence-in-depth because no single technical, institutional or operational safeguard is sufficient.

Who this is for¶

What you'll find here¶

The section is organised into five parts:

Together, these resources provide a simple path: understand possible futures, explore how risks could unfold, use C·A·G·R to structure preparation, then apply it to the responsibilities your sector owns.

Our approach: Containment · Alignment · Governance · Resilience¶

Advanced AI poses two parallel challenges: safety failures (systems that cause harm) and governance gaps (questions of control and accountability that persist even with safe systems). Both require attention, which is why we recommend Australian organisations structure AGI preparedness around four interconnected pillars:

• Containment – Preventing dangerous AI systems from being developed or escaping control
• Alignment – Making AI systems reliably pursue intended goals and remain safe as capabilities scale
• Governance – Laws, institutions and coordination that shape how AI is developed and deployed
• Resilience – Building capacity to withstand and recover from AI-related disruptions

These pillars form layers of defence. Containment aims to prevent dangerous systems from being created or deployed. Alignment seeks to make systems safer by design. Governance provides oversight and coordination. Resilience prepares Australia to withstand disruptions if other layers fail.

Learn more about the C·A·G·R Framework →

How this relates to SafeAI-Aus core resources¶

Common questions¶