Sovereign AI in Heavy Industry: What Actually Applies to You

If your client is an Australian Government agency handling sensitive data

This is as close to a mandate as exists in Australian law right now. The Hosting Certification Framework and IRAP together make offshore AI processing non-viable for sensitive and PROTECTED information. If you're supplying AI-enabled services into that environment, sovereign, on-shore deployment is your baseline — not a preference, not a nice-to-have.

If you're a defence contractor working under DISP

Same conclusion. Security classification requirements and IRAP obligations rule out offshore processing for work touching classified or sensitive material. Sovereign deployment is where you start.

If you're a private sector critical infrastructure operator — energy, water, transport, health

This is where the binary narrative falls apart. Your obligation under the Security of Critical Infrastructure (SOCI) Act is to manage the risk of your AI architecture — not to mandate a specific one. An offshore or cloud-based system is not automatically a problem. What it creates is a supply chain risk that must be documented, mitigated, and approved within your board's risk appetite through your Critical Infrastructure Risk Management Program (CIRMP).

A Tier 1 energy company with enterprise systems centralised offshore and a well-documented CIRMP is more compliant than one with a sovereign cloud setup and no CIRMP at all. The regulator wants evidence that you understand the risk. It does not prescribe the architecture.

If you're a Prime contractor supplying AI-enabled services to a government agency

Your exposure is real but specific. The APS AI Plan requires you to tell your government customer where AI is being used in delivery of your services, and to accept contractual responsibility for that use. Your data residency obligations flow from your client's contract terms — not from a standalone law that applies to you directly. Read your contract. That's where the obligation is.

If you're a sub-contractor in a heavy industry supply chain

Regulatory obligations don't flow to you directly — they flow through your Prime's contract terms. If your head contractor has committed their client to AI governance standards, expect those requirements to start appearing in your sub-contract conditions. They're coming. But right now, they're client-driven, not universally mandated by law.

If you're using AI in safety-critical OT systems — safety monitoring, autonomous equipment, process control

Australia does not currently require formal conformity assessments for high-risk AI systems. That's the European framework. Australia has deliberately taken a different approach — building on existing standards rather than introducing a new mandatory assessment regime. The ACSC guidance on AI in operational technology environments (late 2025) is the practical baseline. Document the risk in your CIRMP and your Safety Management System. Keep an eye on whether Australia moves closer to the EU position — but don't build your compliance program around a requirement that doesn't exist yet.

If you're building or expanding data centre or AI compute infrastructure

This is the population the "sovereign AI" conversation was originally written for. The National AI Plan's Expectations for Data Centres and AI Infrastructure Developers apply directly. National interest, energy, water, workforce, and data sovereignty — all in scope. Alignment with these expectations affects how your project gets prioritised by Commonwealth regulators. Non-alignment makes your approval pathway slower and harder.