Constructing the UN’s technocratic authority

Legitimising their position as actors (re)producing this global emerging SI, the reports open by situating the UN as a central actor in the governance of counterterrorism and new technologies. It is recalled that, under the UN Global CounterTerrorism Strategy, Member States ‘resolved to work with the United Nations’.Footnote ⁷² The Strategy also emphasised that ‘Member States may require assistance [from the UN] to meet these commitments’Footnote ⁷³ – a narrative that presents the UN as a leading provider of expertise, guidance, and capacity-building. Similarly, it is stated that the UNCCT ‘stands ready to support Member States and other counterterrorism partners in countering the threat of AI by terrorists (sic)’Footnote ⁷⁴ with its skills in diagnosing risks and prescribing appropriate responses being remarked – a narrative framing this organ as a neutral, expert body whose authority in steering global responses is taken as given.

Reflecting how technocratic authority is often constructed through specific methodological claims and scientific collectivity,Footnote ⁷⁵ the reports are co-produced by two specialised UN bodies – UNCCT and UNICRI – each recognised as experts in their respective domains. This expert collaboration strengthens the construction of scientific objectivity, something that also emerges in the reports’ methodological signalling. Both documents explain that their findings emerge from ‘desk-based research and open-source information, such as articles, official reports and media reports’ and from an ‘Expert Group Meeting […]’.Footnote ⁷⁶ By foregrounding a methodology based on systematic research and expert consultation, the reports reproduce the construction that the knowledge they provide is scientific, objective, and apolitical.Footnote ⁷⁷

The reports also constitute key technocratic artefacts within this global emerging imagination – and they also construct their own role within it. For example, they claim to ‘serve as an early warning for potential malicious uses and abuses of AI by terrorists’.Footnote ⁷⁸ They describe their own role as ‘to help the global community, industry and governments […] to ensure new technologies are used to bring good and not harm’Footnote ⁷⁹ and ‘to contribute to understanding the potential risk of AI falling into the hands of terrorists’.Footnote ⁸⁰ Therefore, they position themselves as guidance to understand these future possible threats and as leading actors in the formulation of precautionary security.Footnote ⁸¹

The documents frame themselves as global technocratic references within this imagination.Footnote ⁸² They define AI, outline its applications, map future trends, and offer guidance for ‘future’ capacity-building and policy measures.Footnote ⁸³ They articulate rhetorical moves such as ‘Is AI the future of terrorism? As this report indicates, this remains to be seen’Footnote ⁸⁴ which position the reports themselves as authoritative guides for navigating uncertain technological futures. Similarly, the regional report notes that although it is framed as region-specific, five of its six recommendations are ‘universal in character’ and of ‘equal relevance’ beyond the region,Footnote ⁸⁵ thereby extending its technocratic reach. They thus position themselves as key technocratic artefacts and, by doing so, they become key sites in the (re)production of the emerging SI.

AI is also narrated as an object that requires technocratic oversight. It is systematically presented as complex, data-intensive, rapidly evolving, and potentially destabilising. AI is framed as a technical domain that demands scientific and specialised competence – something that the UN can provide.Footnote ⁸⁶ By delivering definitions, classifications, and taxonomies of AI, the UN organs position themselves as leading actors through their complexities. As the article shows, AI becomes a technology whose risks can be anticipated only through expert knowledge and coordinated global standards. Through this framing, the UN’s technocratic authority becomes indispensable to governing a future imagined as threatening but technically manageable, a future that is narrated as inevitable, as illustrated in the following section.

Constructing AI’s inevitability through its everyday presence

A second narrative strand frames AI as an intrinsic component of everyday social life. The reports repeatedly describe ‘The integration of digital technologies into everyday life’,Footnote ⁸⁷ depicting AI as a technology deeply embedded in society and whose application is accelerating ‘at an extraordinary pace’.Footnote ⁸⁸ AI is said to be having ‘a profound impact on our society, from healthcare, agriculture and industry to financial services and education’,Footnote ⁸⁹ a dynamic that is also ‘expanding the market’ for AI-related products with its integration ‘into fields such as medicine, economics, communications, insurance, financing and many other fields’.Footnote ⁹⁰

In this sense, this narrative constructs AI as a core part of social and economic life, constructing it as an everyday and, in turn, indispensable technology. On the one hand, embedding AI into the everyday and constructing it as a technology that is vital to society, the narrative lays the ground for the amplification and social dispersion of future AI-enabled risks – i.e., embedded in the everyday, if used maliciously, AI may pose a serious risk to society’s functioning.Footnote ⁹¹ In this sense, this narrative lays the ground to govern the everyday present in the name of preventing future risks.Footnote ⁹² On the other hand, the documents provide little empirical detail on these current applications. AI’s transformative impact is asserted rather than demonstrated, allowing its significance to appear self-evident. Moreover, narrating it as embedded in the everyday also depoliticises this technology – i.e., AI comes to appear as a neutral, ‘taken-for-granted’ tool, and its implementation is not politically scrutinised.

The examples of AI’s everyday uses – such as autonomous vehicles, search engines, spam filters, smart assistants, targeted advertising – reinforce AI’s ordinariness and presence in society. Moreover, the mentions of the ‘every day’ estimated growth of AI marketsFootnote ⁹³ and statements such as ‘the necessary ingredients for the dawn of a new era of AI are at last on the table’Footnote ⁹⁴ reproduce an image of this technology in an accelerating and inevitable development – thus depoliticising its adoption by normalising it. At the same time, references to AI ‘creating poetry’, ‘proving mathematical theorems’, or ‘predicting judicial decisions’Footnote ⁹⁵ reproduce its construction as increasingly ‘intelligent’ and ‘human’.Footnote ⁹⁶ This narrative strand thus materialises AI as deeply embedded in society, while also simultaneously invoking ambitious visions of its future capacities.

Through this narrative, AI becomes not just ever-present, but unavoidable – reproducing technological determinism.Footnote ⁹⁷ The reports insist that ‘AI is here to stay’Footnote ⁹⁸ and that its ‘whirlwind development and integration […] is projected to continue’.Footnote ⁹⁹ The suggestion that AI is ‘obliging actors in all sectors to rethink how decisions are taken’Footnote ¹⁰⁰ further constructs this trajectory as an unavoidable future – thus also depoliticising this process rather than acknowledging that AI’s development is a political decision.Footnote ¹⁰¹ The use of imperative language – AI is ‘obliging’, ‘demanding’, ‘expanding’ – constructs AI as a force whose advance is beyond political choice. This inevitability narrative foregrounds expert-led global governance as the only rational response – i.e., states and other actors should seek technocratic advice to rethink how decisions are taken.

The everydayness of AI also functions as a security problem. By linking AI to widely used tools – search engines, social media platforms, online assistants, and possible risks are amplified, while AI’s uses are depicted as inevitable. Although the reports explicitly acknowledge that ‘AI has not been used directly by terrorist groups to specifically improve or amplify an attack’,Footnote ¹⁰² they claim that terrorists may have used AI ‘passively, or even unwittingly’Footnote ¹⁰³ simply because AI is inescapable. Here, the absence of evidence becomes reframed as a reason for heightened anticipation – reflecting Aradau and Van Munster’s precautionary logics of anticipation.Footnote ¹⁰⁴ Even explicit uncertainties and unknowns are rendered into potential risks and magnified: not only the narrative introduces the potential terrorist use of AI. It also suggests that even unintentional or unconscious interactions with AI are dangerous, discursively enabling interventions and reinforcing the notion that ‘AI-enabled terrorism’ is inevitable – it may happen even unintentionally, thereby constructing an image of an unpredictable and uncontrollable future risk that, for precaution, needs to be acted upon in the present.Footnote ¹⁰⁵ This is further conveyed by depicting AI’s capabilities, as the next section describes.

Imagining AI’s capabilities and governance challenges

The reports reproduce a dual imaginary of AI as they narrate it simultaneously as a transformative opportunity and a looming threat.Footnote ¹⁰⁶ For example, the reports cite the UN Secretary-General, António Guterres, who, in his 2018 Strategy on New Technologies, stated that ‘[w]hile these technologies hold great promise, they are not risk-free […].Footnote ¹⁰⁷ Similarly, the reports assert that ‘AI embodies this duality perhaps more than any other emerging technology today. While it can bring improvements to many sectors, it also has the potential to obstruct the enjoyment of human rights and fundamental freedoms’Footnote ¹⁰⁸ and that ‘Excitement about the potential for societal advancement with AI is, however, tempered by growing concerns about (its) possible adverse impacts.’Footnote ¹⁰⁹ In this sense, the narrative juxtaposes enthusiasm about AI’s revolutionary potential with caution and preoccupation about the future great risks it may entail.

The narrative reflects a hyperbolic juxtaposition based on the amplification of both opportunities and risks. By elevating AI’s potential to contribute to ‘happier, healthier, wealthier and safer’ societiesFootnote ¹¹⁰ and even to the achievement of the ‘the 2030 Agenda for Sustainable Development, by contributing to end poverty, protect the planet and ensure peace and prosperity for all’,Footnote ¹¹¹ the reports construct AI as a revolutionary and promising global technology. These claims are further reinforced by references to technological progress enhancing ‘political participation, and civic action in both democratic and authoritarian systems’.Footnote ¹¹² However, this celebration is systematically juxtaposed to warnings of misuse: ‘AI can also have a dark side: as a general-purpose technology, AI can, just as equally, be used or misused by malicious actors.’Footnote ¹¹³ Similarly, it is emphasised that ‘the advantages’ of emerging technologies ‘also make them appealing to actors with malicious intent’.Footnote ¹¹⁴ The narrative is self-reinforcing: AI’s (constructed) benefits are what make its (potential) dangers particularly acute. Moreover, the narrative also constructs a moralised duality between legitimate and illegitimate users of technology.Footnote ¹¹⁵ While AI is portrayed as enabling global progress and advancing shared international goals, ‘malicious’ users emerge as aberrant figures who threaten not only security but also the realisation of the broader aspirations of international society. Moreover, by portraying AI as powerful, unpredictable and morally ambivalent, the reports construct the idea that its governance requires careful oversight by expert, ostensibly apolitical institutions – thus legitimising the UN’s technocratic solutionism as the main institution in charge of these global aspirations.Footnote ¹¹⁶

While further discussed below, the moral divide is key to this strand. In fact, the connection between AI and terrorism is established from the outset: the first report is titled ‘Algorithms and Terrorism’ and both documents frame their analyses around the potential malicious use of AI and best practices to respond. Statements such as ‘[AI] is a powerful tool that could conceivably be employed to further or facilitate terrorism’Footnote ¹¹⁷ illustrate how the reports establish AI as intrinsically tied to terrorist threats. The reports describe how ‘AI is a powerful tool that could conceivably be employed to further or facilitate terrorism’Footnote ¹¹⁸ – a construction of the threat that is reinforced by claims such as ‘AI can be extremely dangerous if used with malicious intent’Footnote ¹¹⁹ and by the mentions of ‘the dark side of AI – a side that […] remains underexplored’,Footnote ¹²⁰ i.e., unpredictable and unknown. Here, while the former statements explicitly construct AI as an extreme threat, the latter amplifies this perception by presenting AI’s risks as unknown and underexplored. This framing renders AI into a latent and unpredictable threat, reinforcing the idea that it is difficult to anticipate or control – thus, even more dangerous.

The materialisation of AI as a threat is further developed through the detailed description of four key factors. First, AI’s increasing ‘democratisation’ is presented as lowering the barriers to entry for malicious actors.Footnote ¹²¹ Second, AI’s scalability, i.e., its capacity to be deployed on a large scale and across diverse applications, is framed as enabling attacks to be replicated or expanded across multiple contexts, potentially overwhelming existing counterterrorism capabilities.Footnote ¹²² Third, the reports emphasise an ‘inherent asymmetry’, suggesting that AI gives terrorists a strategic advantage, while counterterrorism actors are constrained by human rights obligations.Footnote ¹²³ This contrast constructs counterterrorism actors as comparatively weak and AI-enabled terrorists as disproportionately empowered. Finally, the reports highlight society’s growing dependence on digital infrastructures, presenting this reliance as increasing vulnerability to AI-enabled attacks.Footnote ¹²⁴ Here, AI is portrayed not only as a tool that can be misused but also as a potential site of attack, intensifying the sense of danger – and also calling for precautionary measures to pre-emptively deal with this risk.Footnote ¹²⁵ This move is sustained by the generation of anxiety, as detailed below.

AI anxiety and the precautionary politics of anticipation

The literature has widely analysed how SIs are also sustained through the mobilisation – and construction – of emotions.Footnote ¹²⁶ Along these lines, Johnson and Verdicchio have illustrated how AI’s SIs often circulate ‘AI anxiety’Footnote ¹²⁷ – i.e., a diffuse sense of unease produced by portraying this technology as at once transformative, opaque and potentially catastrophic. In the case under study, fear and uncertainty operate as narrative intensifiers that heighten concerns about ‘AI-enabled terrorism’ and thus become emotions strengthening the imaginary’s political force.Footnote ¹²⁸ This is observable in the Secretary-General’s remarks that AI and emerging technologies ‘inspire anxiety and even fear’Footnote ¹²⁹ or in the claims that AI is ‘extremely dangerous if used with malicious intent’Footnote ¹³⁰ and that, as a technology, it has a ‘dark side’.Footnote ¹³¹ Such statements cast AI as a source of imminent and potentially uncontrolled danger, and they ascribe specific emotions to the emerging SI. This affective atmosphere is further intensified by dramatic framings of AI-generated disinformation, said to risk triggering an ‘information apocalypse’ and ‘reality apathy’,Footnote ¹³² a discursive move that creates anxiety through the reproduction of this very hyperbolic and catastrophic image of the future.

Anxiety is also reproduced through more technical discursive devices, particularly the invocation of the ‘black box’ problem – the idea that humans cannot fully comprehend how deep learning systems produce their outputs, and thus fully control them.Footnote ¹³³ Discussing the black box further ascribes unpredictability, and reinforces narratives of AI as operating beyond human comprehension or controlFootnote ¹³⁴ – capacities that may render the machine unpredictable in the future.Footnote ¹³⁵ The same happens when, despite recognising the limits of current Artificial Narrow Intelligence systems, the reports debate more speculative forms of AI. Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI) are acknowledged as technologies that ‘so far only exist in science fiction’,Footnote ¹³⁶ yet their hypothetical capabilities are described in vivid terms. AGI is framed as ‘more human-like’ and potentially able to perform ‘any intellectual task that a human being can’,Footnote ¹³⁷ while ASI is portrayed as comprising ‘superintelligent machines’ that could surpass human cognitive capacities and even pose ‘an existential threat to humanity’.Footnote ¹³⁸ Here, perceptions of AI as inherently threatening are amplified and anxiety about the future of AI is discursively mobilised and used to materialise imagined futures in the present.Footnote ¹³⁹

Crucially, this affective landscape underpins a precautionary logic. As Aradau and Van Munster argue, in conditions of uncertainty it is imagined futures rather than empirical evidence that justify intervention.Footnote ¹⁴⁰ The reports construct precisely such a horizon of uncertainty: AI is pervasive, fast-moving, only partially understood, and potentially catastrophic. In this framing, not acting becomes riskier Anxiety thus provides the emotional momentum that than acting pre-emptively.⁴⁰ renders anticipatory and precautionary security both necessary and urgent. All in all, these narratives prepare the discursive terrain upon which AI-enabled terrorism is constructed as a specific, urgent and exceptional threat – the focus of the following section.

Narrating ‘AI-enabled terrorism’ as inevitable

A central narrative in the reports is the depiction of terrorists’ adoption of AI as not merely possible but inevitable. This mirrors the broader imaginary of technological inevitabilityFootnote ¹⁴¹: just as AI’s societal integration is framed as unstoppable, so too is its malicious use. Terrorists are characterised as adaptive actors who ‘have adapted to the new digital paradigms of the 21st Century’,Footnote ¹⁴² a discursive move that constructs their technological adoption as part of an established behavioural pattern. Although the reports occasionally acknowledge the absence of evidence, such caveats are immediately counterbalanced by statements that discard uncertainty. This is observable in the following statement: ‘The lack of evidence of the direct use of AI in terrorism should also not be interpreted as indicating that terrorists are indifferent or disinterested in the technology.’Footnote ¹⁴³ Thus, readers are reminded that terrorism is ‘an evolving threat that should not be underestimated’ and that terrorists have already embraced drones, virtual currencies and social mediaFootnote ¹⁴⁴ – a discursive move that naturalises an expected continuum in terrorist adoption of technology. The logic culminates in the claim that ‘although concrete evidence of terrorists’ interest or intention to use AI has not been found, it is prudent to assume that these groups and individuals are aware of this technology.’Footnote ¹⁴⁵ Here, ‘prudence’ becomes the discursive node that articulates precaution and anticipation, even in the absence of present evidence.Footnote ¹⁴⁶

The reports reinforce inevitability by embedding AI within a long historical arc of terrorist innovation. They recall that terrorist arsenals have expanded ‘from the use of knives and guns to aircraft hijackings’Footnote ¹⁴⁷ and that technological developments have ‘transformed the abilities of terrorist and criminal groups […], rendering them global, rather than local, threats’.Footnote ¹⁴⁸ The accumulation of such examples constructs a narrative in which technology adoption is a consistent feature of terrorism, making AI appear as a readily accessible extension of existing capabilities. The inclusion of mundane technologies such as knives and social media platforms further positions AI as similarly reachable, thereby amplifying the threat and normalising its future misuse.

Anxiety is heightened through depictions of terrorists as strategically positioned to exploit technological gaps: ‘Terrorists have been observed to be early adopters of emerging technologies, which tend to be under-regulated and under-governed, and AI is no exception’.Footnote ¹⁴⁹ This framing represents terrorists as technologically savvy actors who move faster than regulatory frameworks, rendering ‘AI-enabled terrorism’ a looming and potentially ungovernable threat. Urgency is reinforced in passages that ask ‘whether – or perhaps better when – AI will become an instrument in the toolbox of terrorism’.Footnote ¹⁵⁰ The temporal shift from if to when crystallises inevitability and legitimises precautionary ‘AI-enabled counterterrorism’ as the only responsible option. Through these cumulative discursive moves, the reports pre-construct ‘AI-enabled terrorism’ and then materialise the threat, as discussed below.

Materialising imagined ‘AI-enabled terrorism’

A second narrative strand to be discussed is the materialisation of ‘AI-enabled terrorism’ – the transformation of a speculative possibility into a tangible, governable threat. Statements highlighting that the reports’ intention is to address the question of ‘whether AI-enabled terrorism could be a conceivable reality, or if it is little more than mere science fiction’Footnote ¹⁵¹ give the reports a scientific tone that (re)produces the UN’s organs’ technocratic authority analysed above. At the same time, the mention of science fiction is important here because, rather than dismissing the threat, it situates the readers within an already familiar visual and cultural repertoire of dystopian futures.Footnote ¹⁵² This statement is followed by the explanation that, ‘For that purpose, it [chapter 5] presents examples of terrorist groups that have demonstrated interest in AI or related technologies, including in videos using facial recognition or unmanned aerial systems […].’Footnote ¹⁵³ More examples of how terrorist groups have exploited past technologies are also reported. It is described how groups such as ISIL, and Al-Qaida affiliates, or the 2019 Christchurch terrorist have used social media platforms, GPS systems, mobile phones, the internet to enhance their strategic movements or their attacks.Footnote ¹⁵⁴ Here, the mentions of ‘emblematic’ terrorist groups helps readers visualise the threat, as they are possibly acquainted with these groups. Moreover, the juxtaposition of such examples with references to hypothetical future uses bridges the gap between current technological practices and anticipated risks. By invoking technologies such as facial recognition or unmanned aerial systems – tools that are familiar and already embedded in contemporary security discussions – the reports suggest that ‘AI-enabled terrorism’ is not a distant or fantastical scenario but an extension of existing patterns. The speculative future is thus anchored in the present, producing a form of narrative continuity in which imagined threats become expected futures.Footnote ¹⁵⁵

This materialisation is reinforced by titles such as ‘Algorithms and terrorism: the malicious use of artificial intelligence for terrorist purposes’ that construct ‘AI-enabled terrorism’ as a current phenomenon. Similarly, the reports claim to offer ‘an in-depth overview of the present and possible future malicious uses of AI by terrorist groups and individuals’.Footnote ¹⁵⁶ This combination of present-tense formulations with future speculations produces a hybrid temporality in which future possible risks produce current vulnerabilities to be avoided through precautionary measures. The same effect is reproduced by the fact that the reports’ analysis draws on both documented uses of digital technologies and uses that, ‘despite the lack of evidence or literature, could become a future reality’,Footnote ¹⁵⁷ a discursive move that constructs a continuum rendering imagined futures analytically equivalent to observed facts, thereby materialising the (speculative) threat.

Another section, titled ‘AI-enabled terrorism through the looking glass’,Footnote ¹⁵⁸ similarly reinforces this materialisation. The chapter describes ways in which AI might enhance terrorism in the future, but the reference to the ‘looking glass’ returns the image of a section analysing an existing threat. The systematic exploration of future ways in which AI may enhance terrorist operations further materialises the speculative threat, helping readers visualise it. Moreover, the materialisation of the threat is also reinforced by the scientific tone of this discussion, the scientific metrics used to predict risk, and the technical language usedFootnote ¹⁵⁹ that depicts the future as tangible and measurable. Consequently, the threat is also materialised through the deployment of technocratic authority, Footnote ¹⁶⁰ a dynamic further scrutinised below.

Scientifically materialising ‘AI-enabled terrorism’

The materialisation of ‘AI-enabled terrorism’ is further reinforced using technical classifications, scientific assessments, and experts’ opinion. Methodologically, the reports build their assessment ‘on desk-based research and opensource information, such as articles, official reports and media reports’.Footnote ¹⁶¹ Apart from academic analyses, these sources also include MIT, EUROPOL, the RAND Corporation, private cybertech firms, civil society and other UN organs and representative reports. The reports are also grounded in two Expert Group Meetings of 27 participants – representing government, industry, academia, and international and regional bodies – convened by the UNCCT and UNICRI in 2021 and describe focus groups’ results organised by other institutions too. As mentioned above, SIs circulate and are sustained by a collectivity of different voices,Footnote ¹⁶² and this shows that this also is the case for the one under analysis. In fact, the collectivity of technocratic voices in the field of counterterrorism and/or AI reinforces the reports’ technocratic authority and further constructs their scientific – and apolitical – objectivity by embedding this production within a broader epistemic consensus.

The reports also construct the threat as scientifically grounded by organising it into systematic, seemingly objective classifications and taxonomies of how AI might be misused.Footnote ¹⁶³ They also assess terrorists’ ‘intent’ and ‘capability’ as separate, measurable variables, which enable the reports to claim objective conclusions about the likelihood and impact of ‘AI-enabled terrorism’. Through this framing, the speculative threat is rendered technically classifiable, legible and, therefore, governable. Still reflecting how technocratic authority builds on scientific analyses and measurements that render a socio-political phenomenon into a technically measurable and manageable matter, Footnote ¹⁶⁴ the reports include a survey assessing the perceived likelihood of the malicious use of AI by terrorist actors. It is reported that of the 27 participants, 44% considered such use ‘very likely’, 56% ‘somewhat likely’, and, notably, no participant deemed it ‘unlikely’.Footnote ¹⁶⁵ These stats convert a future, speculative risk into a measurable phenomenon. This move further materialises it by framing the future as measurable and, thus, not only as conceivable but as scientifically validated and technically plausible.Footnote ¹⁶⁶ In doing so, the reports ground the construction of the threat in empirical evidence derived from technocratic, expert knowledge. They also reproduce their own technocratic authority by presenting these assessments as objective, scientific facts rather than as situated interpretations shaped by particular institutional and political imaginariesFootnote ¹⁶⁷ – thus legitimising the best practices they produce through their embedding within epistemic consensus.

The reports also reproduce precautionary logics by asking experts to rate the likelihood of fictional future scenarios of ‘AI-enabled terrorism’.Footnote ¹⁶⁸ The likelihood of these imagined cases is translated into numerical scores – the results ranging from 2.8 to 3.8 out of 5.Footnote ¹⁶⁹ This discursive move positions speculation within the realm of scientific plausibility and technically materialises the threat by rendering the future numerically accessible. Moreover, by rendering potential futures measurable (and rating them as likely to happen), the reports materialise the future in the present – and thus legitimise anticipation of these future threats.Footnote ¹⁷⁰ This further reveals how these technocratic actors become co-producers of the imagined threat and of anticipatory measures, while their authority is projected into the future too. Scientific rationality and expert judgement render the speculative both legible and actionable and, overall, imaginable, a process further explored below.

Imagining future ‘AI-enabled terrorism’

Imagining the threat – and the technological response – is an important part of SIs.Footnote ¹⁷¹ As such, some more examples of imagined futures can be recalled here – as they work to amplify the imagined danger and, in turn, to legitimise anticipatory responses. For instance, one report introduces three ‘hypothetical yet conceivable’ scenarios.Footnote ¹⁷² These are described in detail, so as to help the reader visualise the future threat. Although explicitly labelled as ‘fictional’, the chapter title – Unfolding the Terrorist Use of AI – frames them as anticipatory descriptions of an already-emerging reality rather than speculative exercises,Footnote ¹⁷³ as the use of the verb ‘unfolding’ returns the image of a description of an already existing dynamic.

Overall, the scenarios conflate science fiction tropes and real-world references, thus encouraging readers to visualise ‘AI-enabled terrorism’ as both spectacular and imminent. At the same time, the hyperbolic imaginary depicted mobilises emotions in the readerFootnote ¹⁷⁴ and AI anxiety.Footnote ¹⁷⁵ The titles themselves amplify this sensationalism. For instance, Scenario 1 is titled ‘Pandemic Problems – Lantium Government and Vaccine Suffer Deadly Blow in Complex Cyber Attack’ (rated 3.1/5 in likelihood). Drawing directly on the affective residue of COVID-19, the reference to vaccines evokes attacks on vital societal infrastructures, while phrases such as ‘deadly blow’ and ‘complex’ reinforce a sense of vulnerability and loss of control. Similarly, Scenario 2, ‘No Place to Hide – Drone Swarm Kills Famous Activist’ (rated 2.8), invokes dystopian imagery of autonomous, weaponised AI. The title itself reifies the perception of inescapability and constructs the threat as targeting civil society, so reproducing a diffuse threat targeting society. Lastly, Scenario 3, ‘Borders No More – Fake Passports Facilitate Bomb Attack in Capital City’ (rated 3.8), draws on familiar security imaginaries by invoking the figure of the border-crossing terrorist disguised as a refugee.Footnote ¹⁷⁶ The association of ‘fake passports’, ‘borders no more’, and a ‘capital city’ underlines the supposed threat to national integrity and urban safety, evoking crowded, chaotic spaces where control is lost.

All in all, these scenarios produce and reproduce possible futures, rendering them imaginable and, in turn, governable. Reified by technoscientific authority, this imaginary bridges the speculative and the actionable and legitimises the construction of ‘AI-enabled counterterrorism’, discussed in the following section.

Constructing the need for ‘AI-enabled counterterrorism’

The narrative constructions analysed so far of AI and, in turn, ‘AI-enabled terrorism’ as inevitable render ‘AI-enabled counterterrorism’ not only as an unavoidable development but also as necessary and moral requirement. This is observable in claims emphasising that ‘law enforcement and counterterrorism agencies across the globe therefore find themselves being pressed to keep up with the digital transformation’Footnote ¹⁷⁷ or that ‘from a security perspective, the need for law enforcement and counterterrorism agencies to adapt to the digital transformation certainly exists’.Footnote ¹⁷⁸ Overall, these instances show how the construction of an inevitable future of ‘AI-enabled terrorism’ produces the need for ‘AI-enabled counterterrorism’. Moreover, technological change is framed as an external force to which security agencies must respond, rather than as a product of political choice – and thus the implementation of AI in security responses is not only legitimised but also depoliticised. Similarly, the reports emphasise AI’s ‘game-changing potential’ and its ability to ‘enhance effectiveness [and] augment existing capacities’.Footnote ¹⁷⁹ These discursive moves render AI’s adoption as a moral necessity and obligation as it could make the difference when facing ‘AI-enabled terrorism’ – thus adopting AI in security responses becomes almost an ethical obligation.

Moreover, the digitalisation of the response is legitimised by intertwining the present and the future. While imagination of possible threats is projected towards the future, the digitalisation of the response is constructed as needed in the present. It is, for example, highlighted that it is important ‘to prepare for the possible future of AI-enabled terrorism’Footnote ¹⁸⁰ as ‘potential threats involving terrorism and AI that may be on, or just over, the horizon’.Footnote ¹⁸¹ This discursive move constructs an inevitable threat located in a near future and thus justifies and calls for the implementation of AI in counterterrorism. Moreover, this move also constructs digitalisation of counterterrorism as an obliged development, rather than a political decision. Therefore, these examples reveal how precautionary logics projected towards the future are activated to legitimise present ‘AI-enabled counterterrorism’.

This is also the case for claims such as ‘A failure of imagination can have deadly consequences’Footnote ¹⁸² or ‘in the absence of evidence, only through speculation can adequate levels of preparedness be ensured’.Footnote ¹⁸³ Here, speculation and imagination are constructed as security practices and they become nodes in the legitimisation of precautionary logics of acting in the present to govern future (imagined) risks.Footnote ¹⁸⁴ Moreover, these statements show how precautionary security logics convert uncertainty into urgency: imagined catastrophic futures are mobilised as a moral obligation to act in the present,Footnote ¹⁸⁵ thereby legitimising pre-emptive and technologically driven counterterrorism measures even in the absence of evidence, a process that is also reified by constructing AI’s efficiency, as illustrated below.

Constructing AI’s efficiency in counterterrorism

The implementation of ‘AI-enabled counterterrorism’ is also legitimised by narrating AI as efficient and infallible,Footnote ¹⁸⁶ that is, as an indispensable instrument for responding to increasingly complex security threats. The reports construction is not straightforward, and they contain caveats and ethical concerns on this matter – such as possible violations of certain human rights, such as the right to privacy. Some technological limits are also emphasised because ‘given the unpredictability of human behaviour and the current state of technological development, the application of algorithms to predict behaviour at an individual level is likely to remain of very limited value’.Footnote ¹⁸⁷ Nonetheless, these limitations do not really open a space for political reflection; instead, they are reframed as technical challenges to be solved through improved models, greater data availability, or enhanced privacy-preserving techniquesFootnote ¹⁸⁸ – that is, through further AI development. Similarly, ethical concerns raised by ‘human rights experts and civil society organisations’, such as discriminatory impacts or the dangers of unwarranted mass surveillance, are narrated as governance issues that can be mitigated through better anonymisation or pseudonymisationFootnote ¹⁸⁹ – thus depoliticising these matters by rendering them into technical issues to be managed and solved by technocratic experts.Footnote ¹⁹⁰

Furthermore, the acknowledgment of the technological limitations is nuanced by narrating how this technology could enhance counterterrorism with, for example, predictive analysis, identifying red flags of radicalisation, detecting mis- and disinformation, automating content moderation and takedown, countering extremist narratives, and managing large-scale data analysis.Footnote ¹⁹¹ Predictive analysis is, for example, narrated as the ‘Holy Grail’ for security forces’, because it would allow authorities to focus on ‘anticipating future terrorist activities and intervening before an attack occurs’Footnote ¹⁹² and gaining ‘deep insights on the network structure of terrorist groups […] and creating policies aimed at reducing attacks’.Footnote ¹⁹³ These claims further construct the need for ‘AI-enabled counterterrorism’, depoliticising its implementation by rendering it a moral necessity. Framed as such by these UN reports, AI becomes a legitimate source of information, a neutral and rational decision-support tool to guide counterterrorism and even identify a ‘terrorist group’. Moreover, this constructed algorithmic objectivity materialises precautionary politics.Footnote ¹⁹⁴ The narrative renders imagination of future threats into objective (present) knowledge and AI as an objective, neutral and apolitical technology to implement precautionary politics and, thus, to enhance counterterrorism. This construction thus legitimises ‘AI-based counterterrorism’ but also expands technocratic authority within global counterterrorism governance, embedding the assumption that future terrorist threats and formulation of counterterrorism responses can – and should – be governed through technological expertise.

As mentioned above, beyond predictive analytics, AI is portrayed as essential for ‘managing heavy data analysis demands’,Footnote ¹⁹⁵ identifying ‘red flags of radicalisation’, detecting the spread of ‘mis- and disinformation by terrorist organisations’, and automating content moderation and takedown across online platforms,Footnote ¹⁹⁶ among other tasks. By assigning these interpretive and classificatory capacities to algorithms, the reports reproduce the imaginary of AI as an objective, reliable, and technically superior mechanism for making sense of complex social processes, e.g., radicalisation.Footnote ¹⁹⁷ It also converts it into a necessary, objective, and apolitical tool to implement precautionary security – as it is depicted as an authoritative epistemic resource that, through its capacity to produce objective knowledge, will allow authorities to navigate the uncertainty of future risks and materialise imaginations about the future into the present.Footnote ¹⁹⁸ This stabilises a vision of the future where security depends on technocratic expertise and algorithmic intervention, a move that further legitimises ‘AI-enabled counterterrorism’ as also illustrated below.

Legitimising exceptional and precautionary responses

Another narrative strand concerns the legitimisation of exceptional, expansive, and precautionary counterterrorism measures. As seen above, the call for imagination – for anticipating what ‘innovative terrorist groups and individuals’ might do in ‘new and unforeseen ways’Footnote ¹⁹⁹ – constructs the threat as simultaneously plausible and unknowable. Unpredictability, here, becomes a discursive node to then articulate the legitimisation of exceptional action. ‘AI-enabled counterterrorism’ is narrated as an exceptional need because it is ‘incumbent upon those responsible for counterterrorism to stay ahead of the curve’Footnote ²⁰⁰ so as to avoid ‘being caught ‘off-guard’.Footnote ²⁰¹ In this framing, precaution becomes a self-authorising logic: the less is known about the threat, the more an exceptional precautionary response is deemed necessary.

As mentioned above, the normative tensions with human rights are acknowledged in the reports. These documents note ‘wide-ranging legal, political and technical challenges’ for all security agenciesFootnote ²⁰² and recognise that AI raises ‘acute human rights concerns’ in already contentious pre-emptive counterterrorism forms.Footnote ²⁰³ They also emphasise that AI-enabled systems may ‘hamper human rights and fundamental freedoms’,Footnote ²⁰⁴ not only when used ‘beyond what is necessary and proportionate’ but also because of structural technological issues such as bias, surveillance, and large-scale data extraction.Footnote ²⁰⁵ Again, these concerns are not framed as grounds for contestation or restraint of the digitalisation of security policies. Instead, they are narrated as ‘key challenges that authorities must overcome’Footnote ²⁰⁶ – implying that they can do so by following technocratic guidance as this can resolve political frictions. Moreover, the narrative that authorities ‘must overcome’ key challenges suggests that exceptional measures are normalised in the name of precautionary security. Here, statements such as ‘The greater the opportunity, the greater the challenge’Footnote ²⁰⁷ also construct respecting human rights as a challenge, rather than a legal responsibility.

This reconfiguration is made clearer in the emphasis that ‘most human rights are not absolute and can be limited if certain requirements are met.’Footnote ²⁰⁸ Arguing that exceptions relating to privacy, freedom of expression and non-discrimination ‘are allowed when such limitations are legally established […] and are necessary and proportionate […]’,Footnote ²⁰⁹ further normalises the idea that derogation is an expected – even necessary – component of effective (morally needed) ‘AI-enabled counterterrorism’. Precautionary logics, thus, (re)articulate the need for exceptional responses where exceptionality is normalised as an adaptation to technological future threats. Legitimising the exception, existing legal and ethical norms are represented as insufficient and ponderous while AI is narrated as requiring and justifying new forms of exceptional (technocratic) governance. Therefore, imagined technological, exceptional threats generate – potentially exceptional – technological responses, laying the ground for the materialisation of ‘AI-enabled counterterrorism’, discussed in the following section.

Materialising imagined AI-enabled counterterrorism

A final strand of the imaginary concerns the materialisation of AI-enabled counterterrorism. Some of the narratives analysed so far already illustrated materialisation, but some more examples can be discussed here. Overall, the main reports’ goal is to translate imagined counterterrorism responses into concrete tools, applications, and best practices. Already mentioned above, the section Probing the Potential of AI Footnote ²¹⁰ sets out a catalogue of the above-mentioned applications of AI in counterterrorism. This allows readers to visualise where AI could enhance counterterrorism and, thus, materialises it as a security practice, but also legitimises it as a necessity. The report explains that these applications have been selected for their ‘prominence in discourse amongst stakeholders’Footnote ²¹¹ – thus deploying technocratic authority to present these as neutral technical possibilities, and not as political matters that need democratic scrutiny.

As mentioned above, certain applications are further elevated through promissory language. Predictive analytics, for example, is framed as the ‘Holy Grail’ for security forces, capable of helping authorities ‘anticipate future terrorist activities and intervene before an attack occurs’.Footnote ²¹² Such depictions confer almost salvific properties on AI, reinforcing the notion that future security depends on the deployment of increasingly sophisticated technical systems. Therefore, imagined counterterrorism capabilities become stabilised as necessary components of a precautionary security in which threats can be pre-empted before they materialise. Moreover, the application of AI in these spheres is legitimised because of its ‘specific potential […]’,Footnote ²¹³ that is ‘The potential for AI to play a role in countering terrorist narratives online is clear’Footnote ²¹⁴ and ‘The use of AI holds considerable potential to contribute to countering terrorism online.’Footnote ²¹⁵ Constructing AI’s ‘potential’ this way legitimises adopting AI despite limited evidence of effectiveness. By framing this potential as self-evident and technically grounded, the reports depoliticise decisions about AI’s deployment and transform possibility into expectation. In other words, AI’s potential becomes here the node for the implementing precautionary ‘AI-enabled counterterrorism’ in the present.

At the same time, the reports acknowledge significant conceptual and normative tensions – the absence of a universally accepted definition of terrorism, the difficulty of translating individual radicalisation into algorithmic patterns, and the risk that counterterrorism powers may serve political ends.Footnote ²¹⁶ These acknowledgements introduce elements of instability into the imaginary, revealing the shaky epistemic foundations on which technical solutions would rest. Yet rather than unsettling the broader narrative, such caveats are absorbed into it: they appear as challenges to be managed through safeguards, human oversight and improved governance frameworks.Footnote ²¹⁷ In incorporating these tensions, the reports maintain the authority of the SI while signalling that its stabilisation requires ongoing technocratic work.