By
Stuart Kerr, Technology Correspondent, LiveAIWire
Military strategy has always been shaped by the technologies
available to the forces that pursue it. The introduction of gunpowder, the
internal combustion engine, and nuclear weapons each fundamentally altered
the logic of conflict. Artificial intelligence is the latest in this
sequence, and there is broad agreement among defence analysts that its
implications are at least as significant as any of its predecessors. The
integration of AI into military systems is already well advanced: autonomous
weapons, AI-powered intelligence analysis, machine learning logistics, and
algorithmic cyber operations are operational realities in several major
military powers, not projected capabilities.
Intelligence and Surveillance: The Data
Advantage
The most consequential near-term military application of AI is in
intelligence, surveillance, and reconnaissance. Modern conflicts generate
enormous volumes of sensor data: satellite imagery, signals intercepts, drone
feeds, and open-source intelligence from social media and communications
networks. Human analysts cannot process this information at the speed
required for operational decision-making. AI can. Project Maven, the US
Department of Defense programme launched in 2017 to apply computer vision to
drone footage analysis, demonstrated that machine learning could identify
objects of military interest at a rate and accuracy that human analysts could
not match, establishing the operational utility of AI in ISR applications and
prompting significant investment in similar capabilities across NATO
militaries.
Autonomous Weapons: The Lethal Decision Problem
The most ethically fraught dimension of military AI is autonomous
weapons: systems capable of selecting and engaging targets without direct
human control of the lethal decision. The International
Committee of the Red Cross has called for binding international
regulation of autonomous weapons systems, arguing that the decision
to use lethal force must remain with a human who can assess proportionality,
distinguish combatants from civilians, and bear legal responsibility for the
decision. This position reflects the requirements of international
humanitarian law, which demands judgment rather than rule-following.
Critics argue further that the accountability vacuum created by
removing human decision-making from the lethal act creates unacceptable
impunity: if a machine kills a civilian, who is responsible? The programming
decisions that led to the outcome may be buried in millions of lines of code,
made years before the incident by engineers who did not anticipate the
specific scenario.
What This Means for You
For citizens in NATO countries, the military AI programmes
underway are being funded through defence budgets and pursued in your name.
Understanding what these programmes involve, what ethical constraints govern
them, and what accountability mechanisms apply is a matter of democratic
importance. The arms race dynamic in military AI also has economic
implications: AI-related defence procurement represents a growing share of
significantly increased defence budgets across NATO, and the commercial AI
sector and defence AI sector are increasingly intertwined.
As LiveAIWire has covered in analysis of AI
in law enforcement and security, the transparency and
accountability challenges that arise when AI systems make consequential
decisions apply with even greater force in military contexts, where the
consequences include lethal force and the legal frameworks are more
contested.
Cyber Operations: The Invisible Battlefield
AI is transforming cyber warfare in ways that extend far beyond
traditional hacking. Machine learning systems can identify vulnerabilities in
target networks at speeds human operators cannot, generate and adapt malware
in response to defensive countermeasures, and conduct influence operations at
a scale that earlier disinformation campaigns could not achieve. The NATO
Cooperative Cyber Defence Centre of Excellence has documented the
integration of AI into offensive and defensive cyber operations across allied
and adversary militaries. Attribution of AI-assisted cyber attacks is more
difficult than attribution of traditional attacks, reducing the deterrent
effect of response threats.
Influence operations using AI-generated content are a related and
increasingly significant domain. The ability to generate realistic synthetic
media, maintain networks of AI-operated social media accounts, and tailor
influence content to specific demographic targets has lowered the cost of
strategic deception campaigns to a fraction of their previous level, with
visible implications for democratic politics and public trust in
information.
Logistics: The Operational AI Layer
Less dramatic but operationally critical is the application of AI
to military logistics. Supply chain management, maintenance scheduling, fuel
and ammunition forecasting, and medical resource allocation have all been
transformed by machine learning. As LiveAIWire has examined in coverage of
AI
in civilian supply chains, the same predictive analytics that
enable next-day delivery can, in military contexts, determine whether forces
in contact have the supplies needed to sustain operations. The US military’s
Project Convergence programme has demonstrated that AI-assisted logistics can
compress the decision cycle for resupply operations significantly, a
potentially decisive advantage in contested environments.
Governance: Racing Ahead of the Rules
International governance of military AI is at an early stage.
Discussions within the Convention on Certain Conventional Weapons have been
ongoing since 2014 without producing binding regulation. The United States
Political Declaration on Responsible Military Use of Artificial Intelligence
and Autonomy has been endorsed by over 50 countries, but is non-binding and
contains no verification mechanism. Russia and China have not endorsed it.
The consequence is that the rules governing military AI are being determined
primarily by the deployment decisions of major military powers rather than by
international law, giving this governance gap unusual urgency given the pace
of technological development.
The Human Dimension: Soldiers, Commanders, and
Machines
Military AI raises questions not only about international law and
strategic advantage but about the human experience of war. Soldiers who
operate AI-assisted weapons systems, commanders who make decisions on the
basis of AI-generated intelligence assessments, and analysts who review
outputs of automated surveillance systems are all navigating a new
relationship between human judgment and machine recommendation in contexts
where the consequences of error are lethal.
Research on human-machine teaming in military contexts has
identified consistent patterns: operators tend to over-trust automated
systems, particularly when those systems have previously been accurate,
leading to acceptance of system recommendations without adequate critical
assessment. This automation bias can lead to errors in high-stakes situations
that well-calibrated human judgment would have avoided. Military training
programmes are beginning to address this explicitly, developing doctrines and
exercises specifically designed to maintain critical human judgment in
environments where AI recommendations are continuously
available.
The accountability question is equally important at the human
level. When a commander accepts an AI-generated targeting recommendation that
results in civilian casualties, the legal and moral responsibility for that
outcome does not transfer to the machine. Understanding the limitations and
error modes of AI systems is now a core competency for military commanders,
in the same way that understanding the capabilities and limitations of a
specific weapons system has always been. The integration of this understanding
into military education and training is developing, but unevenly across
different armed forces and different levels of command.
Simulation and Wargaming: AI as Strategic
Advisor
One of the less-discussed but practically significant applications
of AI in military strategy is in simulation and wargaming. AI systems can
generate and evaluate millions of simulated conflict scenarios far faster
than human wargaming teams, identifying strategic vulnerabilities, testing
the robustness of operational plans against a range of adversary responses,
and surfacing non-obvious dependencies in complex military systems. The US
and UK militaries have invested substantially in AI-enhanced simulation
environments, and the outputs of these systems increasingly inform strategic
planning at senior levels.
The risk of over-reliance on simulation outputs is analogous to
the automation bias problem in operational contexts. Simulations are only as
good as their underlying models, and military AI simulations are trained on
historical data and designed scenarios that may not adequately represent the
conditions of future conflict. A system that consistently recommends a
particular operational approach because it succeeded in a large number of
simulated scenarios may be training planners to favour an approach that is
well-optimised for the simulated world but vulnerable in the actual one.
Maintaining genuine critical engagement with simulation outputs, rather than
treating them as authoritative strategic guidance, is a professional
responsibility that AI adoption makes more rather than less
important.
About the Author
Stuart Kerr is the Technology Correspondent at LiveAIWire,
covering artificial intelligence across society, policy, and industry. About
LiveAIWire.