Joe Edwards
Published in National Security Journal on 20 October 2026.
DOI: 10.36878/nsj20261020.06
Download full PDF version – Artificial Intelligence, Terrorism, and Counter-Terrorism in New Zealand (578 KB)
Predominantly within the last five years, apprehension over the ambiguous psychosocial
and security ramifications that will likely follow unrestrained developments
in artificial intelligence (AI) has intensified around the world. Drawing on
publicly-released official information, a thematic analysis of internationally published
research, and interviews with subject-matter experts, this article finds that
while New Zealand policymakers do recognise the significance of this evolving
technology, they seemingly pay insufficient attention to the multifaceted relationship
between artificial intelligence and both the problem of terrorism and the
practice of counter-terrorism. This article argues that New Zealand policymakers’
treatment of this relationship seems decidedly limited because they appear
to be focused on reaping immediate economic-productivity benefits while protecting
New Zealanders’ privacy rights. Demonstrating the often unrecognised
and unrealised value of scholarly knowledge while contributing to a growing New
Zealand security studies literature, this article makes several constructive recommendations
for policymakers to ignite and frame a national security conversation
on the risks and opportunities emanating from the rapid developments in
artificial intelligence.
Keywords: New Zealand Security Studies, Violent Extremism, Radicalisation, Terrorist
Mobilisation, Generative AI, Deepfakes, Chatbots, Social Engineering, Disinformation,
Cybersecurity
Introduction
As the science of producing machines that demonstrate human-like logic and decision-
making capabilities, artificial intelligence (AI) is one of the most enigmatic and
widely misunderstood areas within the field of computer science.1 Despite an eruption
of media and public interest over the preceding five years or so, misconception
and anthropomorphic falsehoods perpetuate throughout society and popular culture,
clouding what exactly constitutes AI.2 It was first conceptualised as a field of computer
science in the 1950s when AI pioneer, John McCarthy, described it simply as “the science
and engineering of making intelligent machines.”3 While the basic concept of AI
has not changed much since its inception, public perception and understanding has
altered drastically over the following decades. The field of AI has experienced waves of
hot and cold seasons – that is, summer periods of public hype, strong developmental
growth, and investment cycles are followed by winters where expectations surrounding
the technology largely fell flat – first in the mid-1970s with a second wave occurring
during the 1980s-90s.4 Perhaps humanity is now experiencing the rise of a third such
wave. This time, however, AI hype is driven by tangible and exponential developments
in computer science, with the most striking being recent breakthroughs in big data
accessibility, computational power, and sophisticated deep neural networking design,
which have, in turn, accelerated innovations in machine and deep learning as well as
newfound generative AI (GenAI) capabilities.5
Despite its relative longevity as a scientific field, no consensus exists on the definition
of artificial intelligence.6 This is partly because AI is an amalgam of various detached
narrow-set technologies continually and interdependently developed over decades.7
However, AI can be broadly divided into two tiers: first, machine learning (ML), which
comprises of algorithms that learn from data and identify patterns used to classify or
predict outcomes with improved performance correlating to increased data ingestion;
and secondly, deep learning (DL), which is a subset of machine learning that utilises
multi-layered neural networks capable of processing extensive volumes of data and sequentially providing outputs from identified data patterns. Such systems may include
pre-trained large language models (LLMs), such as ChatGPT, which can sort, condense,
translate, predict, and generate seemingly novel content while maintaining vast textbased
datasets.8 However, the current field is constrained to individual narrow-set applications,
or specific systems which excel at single tasks and problem sets: for instance,
facial recognition, voice, sound and text pattern detection, multi-language translation,
image analysis, predictive modelling, and decision-making, as well as novel content creation.
9 Despite recent media sensationalism, the AI field is nowhere near reaching “generalised”
or “super” (perhaps even sentient) levels and experts question whether this
level of technological advancement is even possible.10 For now, “narrow AI” applications
are fundamentally predicated on human-designed architectures and various training
methodologies largely powered by underlying machine and deep learning models. Such systems are fed with vast datasets, sometimes including exemplar prompts, to identify
patterns and make predictions and decisions – typically through a process of self-supervised
or reinforcement learning, and use of artificial neural networks. This process
convincingly resembles human-like intelligence and intuition, though, in reality, the
resemblance is illusory, rather than familial, because it merely reproduces highly sophisticated
and deep-layered pattern recognition.11
Even recent advances in generative AI (a specialised subset of deep learning) should
not be conflated with human-like consciousness or cognisance because such systems
consist of expensive purpose-built structures, which attempt to artificially imitate human
neurological networks, utilising complex and multi-layered structures of synthetic
nodes.12 In effect, this technological leap enables AI-driven capability to create novel
and increasingly hyper-realistic audio and visual content derived from human prompts
by drawing on vast datasets and associated complex patterns. These generative AI applications
have already become infused into many real-world applications, including
the Metaverse and numerous public platforms (i.e. Microsoft, Google, and OpenAI).13
While generative AI platforms excel at retrieval of scattered information and producing
audio/text/visual responses, these narrow models are prone to erroneous and unpredictable
hallucinations, deeply ingrained human cognitive bias, and privacy and confidentiality
leakages; and they often misleadingly fail to understand nuanced abstract and
contextual ideas that humans inherently can and frequently do.14
While an authoritative definition of AI remains elusive there is an emerging consensus
that, unlike previous winters, AI is no longer merely hype.15 AI proliferation and reliability
will undoubtedly continue to exponentially advance, making it more powerful
and accessible, difficult to control and, for governments, to effectively regulate. The risk
of exploitation by nefarious users and unexpected social consequences is a widespread
security concern.16 Furthermore, there are mounting apprehensions surrounding AI
tools that will undoubtedly present a multiplier effect for unintended and unforeseen
consequences as well as inadvertent zero-day societal dangers portentously manifested
by a mass (unchecked) AI uptake.17 Permitting full AI autonomy in processing volumes
of public data with minimal human oversight is another concern, signalling a range of
as yet unforeseeable black swan outcomes.18
In this article, artificial intelligence is defined broadly as “systems that display intelligent
behaviour by analysing their environment and taking actions, with some degree
of autonomy, to achieve specific goals.”19 The article relies on data collected from
semi-structured interviews conducted with seven subject-matter experts and written
questionnaires completed by them so that it is informed by up-to-date expert opinion
on the phenomenon of AI and how, in particular, New Zealand responds to this technology’s
impact on the problem of terrorism and counter-terrorism practices.20 Interviews
comprised of one-hour, one-on-one meetings and they are hereafter generically referred to as ‘New Zealand experts’ to further preserve their anonymity. The article also captures official information released after formal requests sent to nine government agencies likely have a direct interest in AI. Other research data includes recently published documents, reports, articles and industry-panel webinars pertaining to emergent technologies.
Themes extracted from the interview data were compared to publicly released official
information and an array of other publicly-available documents, including international
case studies, academic journal articles, and government and non-governmental reports.
This analysis yielded important insights. It found that, while New Zealand policymakers
do recognise the significance of this evolving technology, they pay insufficient attention
to the multifaceted relationship between artificial intelligence and both the problem of
terrorism and the practice of counter-terrorism. This article argues that New Zealand
policymaker’s treatment of this relationship seems decidedly limited because they appear
to be largely focused on reaping immediate economic-productivity benefits while
protecting New Zealanders privacy rights.
This introduction has set the scene by problematising the AI phenomenon. The remainder
of the article is divided into three main sections. The first section describes New
Zealand’s response to this complex and dynamic phenomenon. By recommending legal
reform, increased cooperation with international partners, improved multi-agency
collaboration, deeper investment in research, social licence and public education, and
strengthened oversight measures, this section constructively signals key deficiencies in
New Zealand’s current strategic posture that warrant urgent remedy. While contributing
to New Zealand’s growing security studies literature, the second and third sections
demonstrate the value of scholarly knowledge for lawmakers, policymakers, and security
professionals. They do so by examining, first, the risks that AI poses around the
problem of terrorism – including social engineering, disinformation brokerage, chatbot-
enabled recruitment, research and planning as well as force multiplying attacks –
and, second, the opportunities AI presents for counter-terrorism practice. This includes
augmenting analysis and threat detection, enhancing operational and administrative efficiencies, and improving cybersecurity measures. In so doing, the article aims to ignite
and frame a national security conversation on the risks and opportunities emanating
from the rapid developments in the field of artificial intelligence that impact on New
Zealand, including, specifically, its counter-terrorism efforts.
Joe Edwards is a graduate of Massey University and Victoria University of Wellington. He holds a Master of International Security (Distinction) specialising in Counter-Terrorism & Intelligence, a Postgraduate Certificate in Intelligence, a Bachelor of Commerce in Public Policy and Management, and a Bachelor of Arts in Political Science and International Relations. The author acknowledges Dr. John Battersby and Dr. Damien Rogers for their invaluable support and wealth of knowledge provided during the undertaking of this independent research project conducted in 2024 through Massey University, and the National Security Journal. For further information on this research article, please contact j.m.battersby@massey.ac.nz or d.r.rogers@massey.ac.nz.