Australia is racing to build sovereign AI capabilities, but a critical weakness is emerging: the vast majority of companies deploying AI have no way to verify whether their AI systems are making correct decisions. According to research by ServiceNow, just 23 percent of Australian companies have auditing processes to verify AI outputs, despite a 115 percent year-on-year surge in local AI spending. Even more concerning, only 17 percent have systems to manage and track governance and compliance.

Why Does AI Auditing Matter for Sovereign AI?

As Australia pursues its sovereign AI agenda, including the creation of a new Office of AI within the Department of Prime Minister and Cabinet, the lack of auditing infrastructure poses a significant risk. Sovereign AI refers to a nation's ability to develop, deploy, and govern artificial intelligence systems independently, without relying on foreign technology or expertise. Without robust auditing processes, companies cannot ensure their AI systems comply with regulations, avoid bias, or produce reliable outputs. This gap undermines the entire foundation of trustworthy, domestically controlled AI systems.

The Australian government is introducing standards requiring large AI data centres to underwrite new power generation and limit water use, with legislation expected to be introduced in Parliament early next year. However, these infrastructure standards mean little if the AI systems themselves cannot be audited for accuracy and fairness. The governance gap extends beyond technical performance; it reflects a broader challenge in building the institutional capacity needed for sovereign AI.

What Are the Key Governance Gaps Australian Companies Face?

• Auditing Processes: Only 23 percent of Australian companies have established processes to verify that their AI systems produce accurate, reliable outputs, leaving the majority vulnerable to undetected errors or bias.
• Compliance Tracking: Just 17 percent of companies have systems in place to manage and track governance and compliance requirements, making it difficult to demonstrate adherence to emerging AI regulations.
• Rapid Spending Growth Without Controls: The 115 percent year-on-year increase in AI spending has outpaced the development of governance frameworks, creating a situation where companies are deploying AI faster than they can properly oversee it.

This governance deficit is particularly troubling given Australia's broader sovereign AI strategy. The newly established AI Safety Institute is already examining some of the world's most powerful AI models to improve understanding of emerging risks and capabilities. Yet if individual companies cannot audit their own AI systems, the national effort to ensure AI safety becomes fragmented and incomplete.

How to Build AI Auditing Capabilities in Your Organization

• Establish Baseline Auditing Processes: Begin by documenting how your AI systems make decisions, what data they use, and what outputs they produce. Create regular checkpoints to compare AI recommendations against human judgment or ground truth data to catch errors early.
• Implement Governance Tracking Systems: Deploy tools or processes to log all AI model changes, training data updates, and compliance checks. This creates an audit trail that demonstrates your organization is managing AI responsibly and can help satisfy regulatory requirements.
• Assign Clear Accountability: Designate teams or individuals responsible for AI governance and auditing. Without clear ownership, auditing processes often fall through the cracks as companies scale their AI deployments.
• Align with Emerging Standards: Monitor the Australian government's new AI standards and begin preparing your organization to meet them, particularly around data centre power and water use if your company operates large-scale AI infrastructure.

The research from ServiceNow underscores a critical tension in Australia's sovereign AI ambitions. While the government invests in infrastructure, regulatory frameworks, and safety institutes, the private sector is deploying AI at a pace that outstrips its ability to govern it responsibly. This creates a credibility problem for sovereign AI itself. If Australia cannot demonstrate that domestically deployed AI systems are auditable, transparent, and compliant with national standards, the sovereign AI narrative loses force.

The Office of AI, once established, will need to work closely with industry to close this governance gap. Without widespread adoption of auditing and compliance tracking systems, Australia risks building a sovereign AI ecosystem that is technically advanced but institutionally fragile. The next phase of Australia's AI strategy must prioritize not just the deployment of AI, but the governance infrastructure that makes that deployment trustworthy.