Most critical infrastructure risk work starts in the same place. You list the things that could go wrong, you rank them, and you put controls against the ones at the top. It’s a sound way to run a program. It has one blind spot. It can only see the risks that announce themselves.

On 2 November the Critical Infrastructure Security Centre released the fourth edition of its Critical Infrastructure Annual Risk Review [2]. It’s the Commonwealth’s own annual ranking of the risks facing the sectors that keep the lights, the water and the networks on [1]. It ranks them two ways. Plausibility is how likely a risk is, taking account of the threat and the sector’s vulnerability. Damage is the worst case if it happens [1].

The review is careful about its own rankings, and so is this post. It presents them as reference points, and it says operators should assess the risks specific to their own assets [1]. So what follows is the order the review’s text lists. It isn’t a score.

What the government ranks, and in what order

By damage, the worst case, the top five are pre-positioned cyber risk, contracted workforce insider risk, state-sponsored sabotage, AI implementation risk, and limited levels of cyber security [1].

By plausibility, how likely it is, the top five are contracted workforce insider risk, pre-positioned cyber risk, AI implementation risk, third-party cyber risk, and geopolitically driven supply chain disruption [1].

Read the two lists together and a pattern sits in plain sight. Four of those ten entries are about somebody who isn’t you: a contracted worker, a third party, a vendor, a supply chain. Two are about an adversary who’s already inside and doing nothing visible. Only one, limited levels of cyber security, is the thing most programs are actually built to improve. The government’s own order puts people and suppliers above controls.

The thing it fears most leaves no trace

Start with the risk that tops the damage list, ahead of state-sponsored sabotage. Pre-positioned cyber risk [1]. An adversary who’s got in, set up persistence, and is waiting.

A pre-positioned adversary isn’t an event. It’s the absence of one.

That’s what makes it the hardest risk on the list to manage, and it has nothing to do with how good your controls are. A pre-positioned adversary produces no outage. No encryption. No claim, no incident to notify, no lesson to record. There’s nothing to detect, because by design nothing is happening yet.

Now put that next to the document a director signs every year. You can run a full financial year with an intruder resident in your environment. You can review your risk management program, satisfy yourself that it’s current, take the report to the board, have a director approve it, and submit it inside ninety days [4]. Every statement in that report can be true.

The updated CIRMP guidance, which we covered last week, now asks that annual report to demonstrate that the program is current and effective, not just up to date [3]. Effectiveness against a hazard that leaves no trace isn’t a claim you can make from a document. It’s a claim you can only make from looking.

Your program is built to tell you about things that happen. This one is defined by nothing happening.

The other risk at the top is someone you didn’t screen

The risk the review ranks first by plausibility isn’t technical at all. It’s contracted workforce insider risk, and it sits above every purely technical risk on the list [1].

The review is direct about why. Contracted and remote workers hold privileged access. Offshore workforces weaken supervision and raise the exposure to coercion. And the people doing your hiring now face AI-enabled identity fraud, including deepfake credentials and fabricated references that are hard to verify [1]. The review cites an estimate that state-linked insider threats cost Australian businesses upwards of 325 million dollars a year [1].

That’s not a control problem. It’s a question about who’s actually holding your keys, and whether anyone has looked lately. Most security teams were never resourced to answer it, because it’s closer to intelligence work than to engineering.

The same logic runs through the supplier rankings. Third-party cyber risk is fourth by plausibility. Geopolitically driven supply chain disruption is fifth [1]. The review names foreign ownership, control or influence over key vendors as a way an adversary gets persistent access, and it tells operators to map their supply chains beyond direct vendors, work out who really owns and controls them, and diversify [1]. Not to write better questionnaires.

What getting it right has to cover

Read the two lists as one instruction and it isn’t the threat list coming back through a side door. It’s the opposite. The review is telling you where to look, and it’s pointing away from the places most budgets already point.

The five hazard categories the review uses are almost the same five the CIRMP obligation uses: cyber and information, supply chain, physical, natural hazard, and personnel [1]. In most of the operators we see, the cyber chapters are the ones that can be evidenced. There’s tooling, there’s an owner, there’s a budget line, there’s someone who can produce a report. Personnel and supply chain are quieter. Not because anyone’s ignoring them, but because they ask a different kind of question. Who holds access, and was the role ever genuinely assessed for it. Who’s behind a major supplier, what sits behind them in turn, and what happens to your asset when they have a bad week.

AI runs underneath all of it. The review describes AI as having moved from an emerging technology risk in 2023 to a dominant threat multiplier in 2026, embedded across all five hazard categories rather than sitting in one, and it says traditional security frameworks are no longer sufficient against it [1]. That’s its own subject for another day. The point here is narrower. The two risks at the very top of the government’s list are the two that controls alone can’t evidence.

And notice what isn’t near the top. Limited levels of cyber security, the catch-all for whether your technical controls are good enough, is the only purely technical entry on either list, and it sits fifth by damage and off the plausibility top five [1]. The firewalls, the endpoint tooling and the patching that most budgets have gone to for a decade are, AI aside, largely doing their job. The government’s order, and the CIRMP hazards behind it, put the real exposure somewhere those controls were never going to reach: your people and your supply chain.

The question for a board is one sentence

Your program is built to detect and respond to things that happen. So what in it would ever tell you about something that’s deliberately not happening yet, and about a supplier or a contractor whose risk you accepted on their own account of themselves?

Where BayRisk lands on this

Here’s the position, stated plainly.

If the threat is already inside, that’s out of our scope. Finding a resident adversary is detection and response work, and it belongs to your SOC or a partner you scope for it. We don’t do it, and we won’t tell you we do.

What we do is the part the rankings point at. A pre-positioned adversary gets in somewhere, and it stays quiet by hiding in something you already trust. The two most likely ways in are the two risks the government puts at the top: your people and your supply chain. A contractor, a set of credentials, a supplier or their software. Those are the two things we help you manage the risk operation around, and they’re the two most programs have spent the least on.

A program that’s maintained gets attention before a review. A program that’s being run produces evidence as a by-product of normal weeks. Suppliers assessed on evidence rather than on their own account of themselves, including who actually owns and controls them, which is the first thing the review tells you to establish. Access reviewed against roles rather than against a list. And a position on both that you can put in front of a board, a regulator or an insurer on any given day, rather than one you assemble in the fortnight before someone asks.

That’s the work we do, run as a continuous operation on your behalf. The senior judgement, the supplier and personnel work that sits outside what most internal teams were resourced for, and the evidence kept current between assessments rather than built up before them.

The risk itself doesn’t move. It belongs to the board and the executive, by law and by governance, and it can’t be handed to anyone, including us. The annual report is still signed by a director. What changes is whether the person signing it is looking at a position or at a document.

If you want to work through what “current and effective” would mean against the two risks the government ranks highest, that’s a short conversation: hello@bayrisk.com.au.

References

  1. Critical Infrastructure Security Centre, Critical Infrastructure Annual Risk Review, Fourth Edition, November 2026, Commonwealth of Australia. cisc.gov.au
  2. Critical Infrastructure Security Centre, risk assessments page (edition numbering) and upcoming town halls page (2 November 2026 release date). cisc.gov.au
  3. Critical Infrastructure Security Centre, Critical Infrastructure Risk Management Program Guidance, current edition, for the “current and effective” test. cisc.gov.au
  4. Security of Critical Infrastructure Act 2018 (Cth), section 30AG, for the annual report obligation, the ninety-day submission and board approval.