Misreading the Map

Australia’s new defence strategy draws on Ukraine’s experience and the Middle East’s algorithmic wars. It may be reading the wrong lessons from both.

Ukraine is not losing. That fact alone makes it an irresistible reference point for defence planners everywhere. Australia’s 2026 National Defence Strategy (NDS26) cites Ukrainian experience more than a dozen times—in passages on industrial depth, drone warfare, munitions stockpiles, AI-enabled targeting, and the compressed tempo of modern combat. The strategy is right to look but may be wrong about what it sees.

Since NDS26 was finalised, the strategic environment has continued to move. The war in Ukraine has entered a new and grimmer phase. The confrontation between the United States and Iran has demonstrated algorithmic warfare at an operational level previously reserved for wargame scenarios and the damage that drones, dispersed missiles, and ‘mosquito fleet’—can do to strategic choke points, allies and the global economy. Israel’s campaign in Lebanon has extended what was briefly glimpsed in those earlier pager attacks into something closer to a doctrine. And across the Indo-Pacific, the patient, systemic preparation of a great power patiently shaping the environment is continuing without interruption. NDS26 is not wrong to look at these developments. But it is consistently reading them one analytic step too shallow.

None of this diminishes what Ukraine has achieved. Its resistance has been genuinely extraordinary: a society that mobilised against an invasion many thought would last seventy-two hours, sustained a fight across more than a thousand kilometres of front, adapted its military doctrine and its defence industry in near real-time, and maintained democratic governance under existential pressure. Ukraine has earned every lesson attributed to it. The question is whether those lessons travel.

Industry

Start with the industrial argument, treated at length elsewhere in this series. The central point is not about scale but about strategic geometry. Ukraine’s wartime industrial resilience depended on something Australia cannot replicate: a non-blockadeable land border with allied states. Materiel moved continuously through logistics corridors running into Poland, Slovakia, and Romania—supply lines that no adversary could interdict without triggering direct confrontation with NATO. Australia’s supply lines run across thousands of kilometres of ocean, through maritime chokepoints that a capable adversary could contest from the first day of a conflict.

Ukraine’s industrial lesson is about production continuity under attrition. Australia’s strategic problem is logistics under interdiction. These are related but distinct, and the investments they demand are substantially different. NDS26 gestures at this distinction without quite grasping it.

And there is a harder point still. The adversary Australia must reckon with has spent two decades preparing precisely for the interdiction problem—not reacting to it but shaping it. It has mapped maritime routes, seeded sensor networks across the Indo-Pacific, built relationships and bases in the Pacific and Indian Ocean littoral, and established a presence that gives it early warning and early leverage over the sea lanes on which Australia depends. That’s not improvised wartime adaptation; it is patient, systemic preparation by a state that has understood the geometry of Australian vulnerability longer than Australia’s own strategy documents have acknowledged it. The Ukrainian model—rapid adaptation under fire—assumes you reach the moment of conflict without your adversary having already structured the environment. That assumption does not hold in Australia’s case.

Data and AI

The AI argument is more subtle, and more troubling. NDS26 lists artificial intelligence as its first innovation priority. It cites Ukraine’s Griselda targeting platform as evidence that AI can compress decision cycles and deliver decisive tactical advantage. This is accurate as a description of what Griselda did in specific conditions. The question is whether those conditions obtain in the environments Australia is most likely to face.

Griselda worked where dense sensor coverage existed, where ISR—intelligence, surveillance and reconnaissance—was relatively permissive, and where front lines were stable enough to generate the pattern-of-life data that AI targeting requires. The Indo-Pacific theatre—long-range maritime, degraded communications, sparse and unreliable sensor feeds, against an adversary that has studied Western AI-enabled systems, pre-prepared data sources and built countermeasures—is precisely the environment where current-generation AI targeting is most likely to degrade. The technology is not context neutral.

There is a harder question beneath this. Griselda’s effectiveness may reflect not the quality of the AI but the density of the ISR architecture feeding it. If AI battlefield systems succeed primarily where sensor dominance already exists, then the strategic investment priority is ISR architecture, not AI as such. The causal arrow may run opposite to the one the strategy assumes.

Since NDS26 was written, events have added a further, sharper dimension to this argument. The strikes against Iranian leadership in early 2026—in which AI-enabled integration of traffic camera feeds, signals intelligence, human intelligence, and satellite imagery allowed actors to pinpoint the location of senior figures with a precision previously unavailable¹—represent a qualitative step beyond anything seen in Ukraine. This is what has been called ‘algorithmic warfare’: not merely faster targeting but the fusion of disparate data streams into decision quality that no human analyst could achieve at the required speed. It works when the ISR architecture is rich, integrated, and continuously updated.

Israel’s operations in Lebanon extended this logic into infrastructure. The 2024 pager and walkie-talkie attacks were startling because they weaponised the supply chain itself—turning commercial devices into precision munitions through hardware compromise. What followed refined that instinct: cyber access used not for disruption but for persistent surveillance, feeding targeting data into AI-enabled systems, collapsing the distinction between intelligence collection and kinetic effect. The lesson is not simply that AI is powerful. It is that algorithmic warfare is a system of systems, and its effectiveness varies catastrophically with the quality of the underlying data environment. Australia’s ISR architecture across the maritime Indo-Pacific is not ready to support this model. NDS26 does not acknowledge that gap.

Data and Cyber

This matters because the same data-environment logic applies in reverse. The most consequential development in Australian strategic vulnerability over the past three years has not been a weapons system. It has been the documented scope of cyber pre-positioning inside critical infrastructure—the Volt Typhoon and Salt Typhoon campaigns, in Australian systems as in American ones—that represents a coercive capability being assembled now, against a future moment of strategic crisis. The coercive threat is not ‘we will attack your infrastructure’ but ‘in a Taiwan contingency, you will face simultaneous domestic disruption’. That threat is latent, patient, and already inside the wire.

NDS26’s treatment of cyber is notable for what it does not say. The document acknowledges that ‘malign actors—both state and non-state—are continuing to improve their cyber capabilities’. It does not engage with pre-positioning as a distinct and qualitatively different strategic problem from conventional cyber-attack. The difference matters enormously. Pre-positioning is not about stealing data and departing. It is about embedding capability inside critical infrastructure for future use—acquiring the operational knowledge needed to disrupt rather than merely observe, maintaining access for months or years, and activating it at a time of the adversary’s choosing. You cannot health-check your way out of an adversary that is already inside.

The information warfare dimension has developed in parallel, and NDS26 is similarly opaque about it. The Hydra series has flagged the maturation of AI-enabled influence operations: content generated at scale, linguistically fluent, contextually coherent, at negligible cost per unit. The strategic assumption that used to bound influence operations—that the scale of an operation reflects the seriousness of the adversary’s intent—no longer holds. Small operations can now be large. Large operations can now be deniable. A strategy that lists AI as its first innovation priority should have confronted the AI-enabled threat environment as well as the AI-enabled capability aspiration. It did not.

The Adversary has a Say

More worrying is the omission of the adversary’s evolving posture. NDS26 acknowledges that ‘in 2025, China unveiled highly advanced weapons and military hardware’ spanning nuclear-capable air-launched ballistic missiles, hypersonic anti-ship missiles, stealth attack drones, AI-enabled military systems, and robotic platforms. It notes that PLA intercepts of foreign military vessels and aircraft are ‘becoming more frequent and, at times, unsafe and unprofessional.’ These are accurate observations presented in the muted language of a risk register, as though the accumulation of specific, documented capabilities constitutes a background condition to be noted rather than a strategic challenge to be directly addressed.

That framing understates what is actually happening. China has been watching the Ukraine war and the Middle East conflicts with the forensic attention of a practitioner. The lessons it is drawing are not the same lessons that appear in NDS26. It has watched algorithmic warfare demonstrate that ISR density is the decisive input, and it has invested accordingly. It has watched Western democracies struggle to sustain industrial production at wartime tempo, and it has pre-positioned accordingly. It has watched the United States commit forces to multiple theatres simultaneously, and it has noted the simultaneity problem that the US NDS26 itself candidly describes. Its navy now has more hulls than any other in the world. Its shipbuilding capacity dwarfs America’s. Its missile inventory—DF-26 ‘carrier killers’ among them—has been designed expressly to render surface fleets vulnerable in the western Pacific.

The Taiwan question sits at the centre of this posture, and NDS26 does not engage with it at the level of strategic consequence it deserves. The document notes ‘flashpoints in the Taiwan Strait’ among a list of regional risks. This is the cartographic equivalent of describing a fault line as a feature of local topography. Taiwan is not a flashpoint among others. It is the contingency around which the adversary’s entire military modernisation programme has been organised, and around which Australia’s alliance obligations, AUKUS commitments, and geographic vulnerability would be most acutely tested simultaneously. A strategy that cannot say this plainly is a strategy that has not fully come to grips with its own strategic environment.

What a Taiwan contingency would mean for Australia is not speculative. It would mean US forces prioritising the immediate theatre. It would mean contested maritime chokepoints. It would mean the supply chains that NDS26 describes as needing ‘resilience’ being broken, not simply stressed as though COVID or the Iran conflict. It would mean Volt Typhoon-style latent capabilities activated against domestic infrastructure at a moment of maximum strategic pressure. And it would mean Australia needing to make decisions—about force commitment, about logistics, about democratic authorisation of action—at a speed and under conditions that its institutional architecture is not currently designed to handle.

Governance and Democracy

This brings us to the most significant omission in NDS26. A document that lists AI as innovation priority number one contains no meaningful treatment of human-machine teaming policy, no autonomous weapons doctrine, no framework for democratic accountability when targeting cycles are compressed below the timescales of human deliberation.

This silence is not a minor gap. When a strategy is enthusiastic about acquiring a capability but silent about governing it, the most likely explanation is that the institutional capacity to govern it does not yet exist. A capability without governance architecture is not a strategy. It is a procurement programme.

The governance question is constitutional, in the fullest sense. Democratic accountability over the use of lethal force depends on chains of authority that humans can trace and decisions that humans can contest.²^. AI systems operating at machine speed within opaque decision architectures do not eliminate that requirement. They make it harder to satisfy. The algorithmic warfare demonstrated in the Iran strikes and the Lebanon operations was conducted by states with specific legal frameworks, intelligence oversight mechanisms, and command accountability structures. Australia has none of these calibrated for AI-enabled operations. If it acquires AI-enabled targeting without simultaneously building the governance framework to keep human authority real rather than nominal, it will have created capabilities that undermine the democratic principles the strategy claims to defend.

The Iran and Lebanon cases add a further dimension that Australian strategy has not absorbed. Both demonstrated that algorithmic warfare is not purely a military instrument. It fuses intelligence collection, cyber operations, electronic warfare, and kinetic effect into a single decision cycle. The institutional boundaries between these activities—intelligence agencies, cyber commands, military operations—are legally and organisationally distinct in Australia. Adversaries do not organise by those distinctions. The governance architecture needs to close that gap before the capability is acquired, not after. NDS26 proceeds as if the gap does not exist.

Lost in Translation

Ukraine has given the world extraordinary evidence about modern warfare, and Ukrainians have paid for that evidence in ways that demand both respect and active support. The Middle East has provided a glimpse of where algorithmic warfare leads when ISR architecture matures. Both are genuinely instructive. The question is the quality of the instruction Australia is drawing from them.

NDS26 is strong on aspiration and uneven on diagnosis. The risk is confusing ‘suggestive’ with ‘transferable.’ The Ukrainian experience is suggestive at the level of principle—industrial capacity matters, rapid adaptation wins, AI compresses cycles. The Middle East experience is suggestive at the level of method—algorithmic warfare works when the data environment is ready, and the distinction between intelligence and kinetic effect is dissolving. But the mechanisms depend on conditions that Australia’s geography, its adversary’s preparation, and its own institutional inheritance do not replicate.

A strategy that borrows the surface form of these lessons without doing the translation work is not wrong about the principle. It is wrong about the problem. And getting the problem wrong is how defence planning produces capabilities optimised for the last war, in the wrong theatre, against an adversary that has been quietly preparing for this one—and watching every development in Ukraine and the Middle East with, in all probability, exactly the analytic discipline that NDS26 lacks.

1. Or not, as the strike on the 28 February strike on Shajareh Tayyebeh girls’ school in Minab, southern Iran, showed.

2. Risa Brooks, “The civil-military implications of emerging technology,” Reconsidering American civil-military relations: The military, society, politics, and modern war 221 (2020).