As Australia pushes for sovereign AI, four challenges stand in the way

As AI forces a fundamental shift in the scale, sensitivity and complexity of data both generated and used, Australian regulators face heightened pressure to ensure legislation evolves at pace with industry.

In sectors such as energy and banking, AI is already embedded in critical functions such as credit assessments, fraud detection, energy flow optimisation and supporting policy decisions. As such, regulators are placing greater emphasis on governance and risk controls, focusing on transparency around how AI is deployed, who has access to data and how it is handled across borders.

Challenge 1: Global AI access versus sovereignty risk

Accessing global AI capability often means accepting a degree of external dependency. When AI models, data and supporting infrastructure are hosted offshore, organisations gain the advantages of scale and capability, while also taking on cybersecurity and jurisdictional risks that must be actively managed. Organisations dealing in offshore AI should ask themselves questions such as: What foreign laws apply? When could foreign governments compel access to systems or metadata?

These considerations are especially pressing amid heightened geopolitical tension.

Challenge 2: Physical and infrastructure constraints

Sovereign AI ambitions depend not only on software capability but on physical infrastructure. Power availability, communications latency and the economics of supporting large-scale compute domestically all play a role.

The Australian Energy Market Operator (AEMO) now treats data centres as a standalone demand category. In FY2025, data centres accounted for an estimated 2.2% of National Electricity Market (NEM) demand. By 2029–30, that number is forecast to rise to around 6%. As energy use increases, questions about AI capability have evolved beyond the software and now extend to grid, planning and industrial policy.

Network resilience is another factor. Currently, submarine cables carry over 99% of international data traffic, yet repair capacity is limited and demand continues to grow as AI and cloud usage expand. Dependence on global infrastructure can introduce continuity risks that sit outside the direct control of individual organisations.

Sovereign AI is also connected to broader industrial strategy. Critical minerals and semiconductor supply chains will shape the infrastructure required to support domestic capability. While Australia has a strong position in rare earths and critical materials — essential inputs for semiconductors and AI systems — translating this advantage into end-to-end AI capability will require coordination across multiple layers of the economy.

Challenge 3: Economic trade-offs across the AI stack

Implementing sovereign AI isn’t just a technical decision, it’s an economic one. Value is created at different points across the full AI stack, from the underlying infrastructure (like energy and data centres) through to AI models and the industry-specific applications that use them. However, not all parts are equally profitable.

Today, margins sit disproportionately in infrastructure. Over time, they are likely to shift towards enterprise applications and workflow layers. Over-investing too heavily in a single part of the stack risks limiting long-term value capture.

Another consideration for organisations is how to balance the cost of building and operating local capability against the efficiency and scale of external providers.

Even when workloads are kept onshore, sovereignty can still be undermined by architectural decisions. Data and processing may still flow across borders via analytics tools, support services, logging or failover processes if these dependencies are not carefully managed.

Challenge 4: Talent and capability gaps

Workforce capability is another constraint. Building sovereign AI requires not only infrastructure but also specialised skills, talent retention and the ability to translate technical capability into real-world applications.

Globally, demand is strongest for applied roles such as machine learning engineers, data engineers and product leaders who can deploy AI systems at scale. These skills are in short supply, with employers citing both a lack of qualified candidates and rising salary pressures.

The Australian Government has identified domestic AI capability, workforce uplift and smart infrastructure as national priorities, recognising their role in supporting economic growth and job creation. While initiatives like the AI plan for the Australian Public Service released late last year signal progress in lifting the sector’s maturity, closing the execution gap will take sustained investment.

Sovereign AI as a strategic challenge

The risks associated with limited control over AI systems are not confined to cybersecurity. They include dependency on external roadmaps, exposure to changing commercial terms and the possibility of critical capabilities being degraded or withdrawn with limited notice.

Sovereign AI therefore presents a complex set of trade-offs, spanning compliance, technology and broader strategic considerations that sit at the intersection of regulation, infrastructure, economics and industrial policy.

Rather than being a binary choice between onshore and offshore systems, organisations must make a series of design decisions across the AI stack that balance control, cost and access to global innovation. For boards and CTOs, this means focusing on three priorities.

The first is understanding where control sits today — across data, infrastructure, models and operational dependencies. Second, it is necessary to identify which AI use cases are strategically critical, and therefore require greater control, resilience and transparency.

Third, hybrid architectures need to be designed that balance access to global innovation with selective ownership of key capabilities.

Organisations that take this structured approach will be better positioned to manage risk, control costs and build differentiated AI capability over time.

*Gavin Seewooruttun is the Vice President and Data & AI Lead for Australia at Publicis Sapient. Based in Melbourne, he is responsible for spearheading the company’s data and artificial intelligence capabilities.

Image credit: iStock.com/BlackJack3D