Australia’s AI Warfare Test: When Machine Speed Meets Human Judgement

The Australian Defence Force is moving rapidly towards an AI-enabled battlespace. However, the real strategic challenge may not be whether machines can make decisions faster ... it is whether humans can retain meaningful judgement when they do.

The Australian Defence Force (ADF) is entering a new phase of military technology development in which artificial intelligence, autonomous systems and human-machine teaming are becoming increasingly important to how Australia intends to fight.

An ABC News investigation has highlighted the issue through two of Australia's most visible autonomous capabilities: the MQ-28A Ghost Bat collaborative combat aircraft and the Ghost Shark extra-large autonomous undersea vehicle.

The significance goes well beyond the platforms themselves.

The more consequential question is emerging further upstream: what happens when artificial intelligence begins compressing the intelligence-to-decision cycle faster than human commanders can comfortably absorb it?

That is where Australia's AI transformation becomes a question of military decision advantage ... not simply automation.

Ghost Bat: autonomous does not mean unconstrained

The most striking development is the Ghost Bat's transition from technology demonstrator towards an operational combat capability.

In December 2025, an Australian-designed and manufactured MQ-28A successfully launched an AIM-120 AMRAAM against an airborne target at the Woomera Test Range in South Australia. The aircraft operated alongside an RAAF E-7A Wedgetail and F/A-18F Super Hornet, destroying an Australian-made Phoenix target. Defence subsequently announced approximately $1.4 billion to advance the Ghost Bat towards an operational warfighting capability.

That achievement is important because the Ghost Bat is not remotely piloted in the conventional sense.

There is no fighter pilot sitting inside the aircraft and no drone operator continuously flying it from a ground station.

Instead, it is designed as a Collaborative Combat Aircraft (CCA) capable of operating alongside crewed aircraft and executing elements of a mission autonomously.

But there is an important distinction.

Air Marshal Stephen Chappell, Chief of Air Force, told ABC News that while AI was involved in developing the Ghost Bat, the aircraft itself operates using deterministic programming. Extensive digital-twin testing generates and validates possible courses of action, from which the aircraft selects an approved option.

During the December 2025 live-fire test, an operator aboard the E-7A Wedgetail retained command authority and authorised the Ghost Bat to engage.

That distinction matters.

The current Australian model is not simply:

AI identifies target → AI decides → AI fires.

It is closer to:

sensors and crewed platforms establish the tactical picture → human command authority authorises engagement → autonomous system executes within defined parameters.

That is a very different proposition from an unconstrained autonomous weapon selecting and engaging human targets independently.

Ghost Shark takes autonomy beneath the waves

The same transformation is occurring underwater.

Australia's Ghost Shark is an extra-large autonomous undersea vehicle designed to conduct intelligence, surveillance, reconnaissance and strike missions. In September 2025, the Australian Government announced a $1.7 billion acquisition programme with Anduril Australia covering delivery, maintenance and continued development.

By April 2026, the Royal Australian Navy had formally established its Maritime Autonomous Systems Unit (MASU) to accelerate the operational employment of autonomous maritime systems.

The unit's remit extends beyond Ghost Shark to other uncrewed systems, with the stated objective of integrating persistent, long-range ISR and strike capabilities into Navy operations.

This is strategically significant.

Australia is not treating autonomous systems as isolated technology experiments. They are increasingly being incorporated into force design, doctrine, experimentation, training and operational concepts.

Defence's 2026 Innovation, Science and Technology Strategy reinforces the direction, identifying autonomous systems, artificial intelligence and undersea warfare among six priority technology areas.

The real AI revolution may happen before the trigger is pulled

The Ghost Bat attracts attention because it is visible.

The potentially larger transformation is less visible.

AI can increasingly support the activities that occur before a weapons system is employed: intelligence fusion, surveillance analysis, pattern recognition, target development, battle-management support, course-of-action generation and decision support.

Australian National University's Aina Turillazzi told ABC News that much of military AI operates “upstream” of weapons, helping assemble the operational picture that humans subsequently use to make decisions.

This creates a different category of risk.

A human may technically remain “in the loop”, yet the quality of the human decision can still be affected by the machine's recommendation.

That phenomenon is commonly described as automation bias: the tendency for humans, particularly under pressure, to place excessive confidence in automated recommendations.

The danger is therefore not necessarily that an AI system suddenly takes control.

It may be that humans gradually stop challenging the machine.

Speed can create decision advantage, and decision risk

Military organisations have always sought to accelerate the observe–orient–decide–act cycle.

AI potentially changes the equation because machines can process enormous quantities of information at speeds that human staffs cannot match.

That creates an obvious military advantage.

But speed is not synonymous with accuracy.

If an AI-enabled system identifies potential targets faster than commanders can interrogate the underlying evidence, the organisation can become trapped between two competing imperatives:

move faster than the adversary, or slow down sufficiently to understand what the machine is telling you.

That tension is becoming increasingly visible in contemporary warfare.

The International Institute for Strategic Studies has examined the expanding use of AI-enabled military technologies in the Middle East, including AI-enhanced ISR, battle management and decision-support systems. Its analysis also highlights concerns surrounding the use of systems associated with Israeli targeting operations in Gaza.

These cases should not, however, be treated as direct equivalents of Australia's Ghost Bat architecture.

The technologies, command arrangements, operational environments and stated safeguards differ considerably.

That distinction is important when discussing AI warfare.

Ukraine is demonstrating the speed of the adaptation cycle

Ukraine provides another important reference point.

Uncrewed systems have become central to battlefield adaptation, with both sides rapidly modifying drones, sensors, communications systems and software in response to operational experience.

Australia's Chief of the Defence Force has explicitly cited Ukraine as evidence of the growing importance of autonomous and uncrewed systems on the modern battlefield.

The lesson for Australia is not simply that Defence needs more drones.

It is that software-defined warfare compresses the capability-development cycle.

A platform can be modified, tested and redeployed faster than traditional military acquisition models were designed to accommodate.

AI potentially accelerates that cycle again.

The result is a battlefield in which the decisive advantage may increasingly belong to the force that can integrate data, software, autonomy, sensors and human command faster than its opponent.

The human safeguard is therefore more important, not less

Australia's existing Defence AI policy framework explicitly retains human accountability.

Defence states that human judgement and accountability are central to the lawful, legitimate and responsible use of AI, with designated personnel remaining accountable for AI-enabled decisions and outcomes.

This is particularly important where lethal force is concerned.

The international debate is moving in the same direction.

In August 2026, United Nations Secretary-General António Guterres and International Committee of the Red Cross President Mirjana Spoljaric renewed their call for international rules governing autonomous weapons.

Their appeal specifically highlighted concerns about unpredictable autonomous weapons and systems that target humans, arguing that increasingly autonomous systems could reduce meaningful human control over the use of force.

The ICRC and UN position is therefore not simply “ban AI”.

It is about establishing boundaries around where machine autonomy should stop and human responsibility must remain.

The Petrov lesson for the AI age

One of the most useful historical parallels raised in the ABC investigation is the 1983 Soviet false-alarm incident involving Lieutenant Colonel Stanislav Petrov.

Soviet early-warning systems indicated that the United States had launched missiles.

Petrov judged that the warning was likely false and chose not to treat the automated alert as definitive.

The system was wrong.

The lesson for an AI-enabled military is not that humans are inherently superior to machines.

It is that a human decision-maker must retain the capacity, authority and confidence to challenge the machine.

That is a much higher standard than simply keeping a person somewhere in the chain of command.

A nominal human “on the loop” is of limited value if operational tempo, interface design, organisational culture or overwhelming confidence in AI recommendations makes intervention practically impossible.

From automation to military decision advantage

This is where Australia's AI transformation becomes strategically interesting.

The objective should not be to create a “silicon commander” that replaces human command.

The more useful model is a human-machine command architecture in which:

• AI processes information at machine speed;
• autonomous platforms provide persistence, mass and reach;
• humans establish intent, constraints and accountability;
• AI exposes uncertainty rather than hiding it;
• commanders can interrogate machine recommendations;
• lethal decisions remain subject to meaningful human control; and
• the system deliberately creates opportunities to pause when the evidence is incomplete.

The emerging contest is therefore not simply human versus machine.

It is human-machine teams versus human-machine teams.

That changes the definition of military decision advantage.

The winning force may not be the one with the most sophisticated AI model.

It may be the force that best integrates AI speed with human judgement.

Australia's Ghost Bat and Ghost Shark programmes demonstrate that this transformation is no longer theoretical. The platforms are moving from experimentation towards operational capability, while Defence is simultaneously developing the policy, command-and-control and workforce structures required to employ autonomous systems.

The strategic challenge now is ensuring that technological acceleration does not outrun institutional judgement.

Because in warfare, the hardest decision may not be getting a machine to act.

It may be knowing when the machine should be told to wait.

Follow @NovationemForum for daily business, financial markets, geopolitics & AI analysis.