AI is radically transforming how National Security Intelligence is collected, processed, analyzed, and disseminated, although not in the same way for each of these steps. Drawing from Intelligence Theory and the concept of Revolution in Intelligence affairs, this paper aims to understand what factors are most likely to evolve in the nearer future, based on the concept of a "weak" AI capability (not fully sentient). It concludes that AI is going to greatly enhance the velocity and scale of intelligence through automation. On the other hand, rivals might be more inclined to disrupt and deceive each other's intelligence processes, as machines increasingly spy on other machines. Strategic evolutions of the use of AI will fuel discussions on doctrine and ethics.

Introduction: AI as a Driver of Change in Intelligence

The simple purpose of Intelligence as a practice is to transform data into actionable knowledge. However, this is certainly not straightforward: "Data, absent rigorous and theoretically based analysis, remain simply data or information."1

The technological evolution the world is undergoing is both transforming the way we do these tasks and making them more indispensable than ever.  This  "has had a major impact on intelligence practice." 2 Intelligence agencies have always been savvy of new technologies, as the analysis of data plays a central role for them – it has therefore evolved significantly in the past according to Brantly (2018).3  However the argument of Gannon (2008) is that the information age has brought intelligence to "warp speed", both through the ubiquity of data there is to collect today – even toasters and coffee machines could be sources of intelligence today4 – as well as the pace at which innovations advance.5  Artificial Intelligence (AI) is the best example of both of these factors, finding concrete applications in the Intelligence community (IC) throughout the Intelligence Cycle6  for years now, to the point of Big Data being characterized as a megatrend in global intelligence by Kojm (2016).7  In fact, AI is often characterized as "our generation's race to the moon".8 Its implementation is inevitable.9 How the technologies will be applied to the Intelligence Communities is an open question. The key therefore is more about integration of innovations than about how to innovate more.10 A solid doctrine on AI in Defense in Western Countries is currently lacking.11 Therefore, this is not a purely technical subject – but rather, a strategic one.12 The ever-faster pace of innovation and its effect on the whole of society present an opportunity to analyze the impact of AI on the Intelligence Community. The following research question will thus guide this paper: How does AI impact the individual components of the Intelligence Cycle and how can the threats and opportunities arising from increasingly relying on machines be addressed efficiently?

Research Question & Methodology

This paper will aim to identify the transformations induced by the use of AI throughout the Intelligence Cycle. It will therefore start with a definition of the Cycle, followed by the introduction of the concept of the "Revolution in Intelligence Affairs" (RIA), which guides the argument. The five components of the Intelligence Cycle will then be analyzed separately. The analyzed sources will be industry reports as well as secondary literature. After that, the paper will assess the strengths, weaknesses, opportunities, and threats of the use of AI by the Intelligence Communities through a SWOT analysis, based on the findings of the first part. A SWOT helps to find an "organization's strengths and weaknesses (S-W), as well as broader opportunities and threats (O-T). Developing a fuller awareness of the situation helps with both strategic planning and decision-making".13

The goal of this paper is to understand the extent to which AI will transform the Intelligence Community to grasp how the RIA is playing out. As the paper focuses itself only on the intelligence cycle, i.e. from collection to analysis, we will limit ourselves to this. Therefore, this paper doesn't discuss possible uses of AI for missions that do not focus on the handling of intelligence. This distinction is important, because the tasks that intelligence services will do are usually broader than solely handling intelligence. They might comprise anything related to national security, such as targeted killings or missions to gain influence, or to damage a rival.

Throughout this paper, AI is defined as "a branch of computer science that deals with developing hardware and software systems that operate autonomously, perform tasks or discern solutions to complex problems in a human-like fashion, and mirror natural intelligence by emulating neurobiological processes and functions".14 For reasons of feasibility, the concept of AI is strictly kept to Hildt's (2019) concept of "weak AI": "weak AI assumes that machines do not have consciousness, mind and sentience but only simulate thought and understanding".15

According to Baker, "AI is hard to define because it draws on a wide spectrum of technologies, subfields, and capacities that make a crisp and singular definition difficult".  As several technological branches of AI are relevant to this paper, we will briefly define them here:

Big data and the Internet of Things (IoT)

"The internet and subsequently the Internet of Things (IoT)[…] has resulted in an explosion of data, metadata, and stored data."


"The revolution in computing capacity is matched by the development in software and algorithmic reasoning, which has transformed the capacity of engineers to search stored data. An algorithm is a set of instructions or calculations to perform a task, often in the form of a mathematical formula. A program is an algorithm that completes a task. 4 Among other things, algorithms are used to search for and detect patterns in data and metadata."  

Robotics and Autonomous Systems

"AI has benefited from the parallel development of civil, military, and commercial autonomous architectures dependent on AI – enabled computers, such as robotics, drones, and smart grids."  

Machine Learning (ML)

"The capacity of a computer using algorithms, calculation, and data to learn to better perform programmed tasks and thus optimize function. […] Deep learning is a machine learning method [that] connects different layers and segments of data internally in a "neural network." […] Deep learning is often compared to how the human brain sorts and connects information through neural brain networks." 

The Intelligence Cycle: Then and Now

The concept of Intelligence cycle most probably emerged just after the Second World War, according to Warner.21 It is "almost a theological concept" of Intelligence Theory, in order to understand how intelligence works – every intelligence officer learns it.22

The Intelligence Cycle usually starts with data collection (of which the types of intelligence collection have been compiled into a table in the annex). The information is then processed (compiled, cleaned, decoded, translated, sorted, etc.) and analyzed (the stage in which data becomes intelligence) in order to produce actionable information. After the analysis and production phase, intelligence is then disseminated (distributed to relevant intelligence customers, usually superiors), which feeds the planning and direction phase. As this is a cycle, the process can then start anew. 23

However, it is also a flawed description of reality, as it can be inaccurate 24: the chaotic nature events often take differ from the perfect case-study intelligence cycle.

"There are no firm boundaries delineating where each operation within the intelligence process begins or ends. Intelligence operations are not sequential; rather, they are nearly simultaneous. Additionally, not all operations necessarily continue throughout the entire intelligence process. The increased tempo of military operations requires an unimpeded flow of automatically processed and exploited data that is both timely and relevant to the commander's needs."25

It is therefore not surprising to see that the Intelligence Cycle has today been relegated to the Appendix of the US Army Manual on Intelligence. The digital revolution is one of the major reasons for this.26  However, it is still very useful as to frame our theoretical approach, which is why we use it.

The goal is to assess how AI processes play out in the different steps of the cycle, as much as how it will affect the relevance of the concept of Intelligence Cycle. It is therefore both viewed as a guiding theoretical concept and as an object of study. Because AI will likely speed up some aspects of intelligence to unimaginable velocities (even nanoseconds, in a cyberattack for example).27 If AI is allowed to automatically act or respond to attacks, the Intelligence cycle cannot be carried out in a traditional way. We can therefore imagine that in some cases, the Intelligence cycle will be blurred, stretched thin, happening almost simultaneously. AI therefore has potential for radical transformation.

Figure 1: The intelligence cycle, as pictured in Lowenthal, 2014

Origins of the Revolution in Intelligence Affairs

The Revolution in Intelligence Affairs (RIA) is a derivative of the concept of the Revolution in Military Affairs (RMA), a term developed by Krepinevich (1992)28  and Gray (2002)29 amongst others. According to Denécé, the "concept was born of technological, political, social, and economic changes that were to fundamentally alter the future of warfare, introducing a completely new type of military and organizational structure for the effective projection of force".30 Similarly to the RMA, the concept of RIA is the claim that technological evolutions can eventually lead to a profound change in the modus operandi of intelligence, as coined by Denécé (2014). The author argues that the transformations in the intelligence community were less apparent – and less discussed – but groundbreaking. He identifies a distinct leap from intelligence in the 90s in comparison to post-9/11 espionage, that started relying heavily on digital means.31 The term RIA has been reused by Vinci (2020) in a recent article,32 in which he makes the argument that AI will alter how "secrets are collected, analyzed, and disseminated" fundamentally – possibly even targeting other machines and not solely focusing on humans anymore. According to him, "through the coming RIA, machines will become more than just tools for information collection and analysis. They will become intelligence consumers, decision-makers, and even targets of other machine intelligence operations." 33

Impact of AI on the Intelligence Cycle

1. Collection

Digital content is growing exponentially: 90 percent of the current total data was created in the last two years.34  It is therefore vital for the intelligence community to bet on digital tools in order to find better, more ways to collect and store data, which is why it turned to big data, machine learning and AI. However, the number of data types collected has also massively increased: because rivals are aware that they could be spied on, encrypted communication is the norm. This means that the Intelligence Community has to use other means to understand what enemies are up to.35 Because everyone has a digital footprint, Intelligence officers collect more data types than in the past. This also means that AI will be much more heavily used in the technical "INTs".  Following benefits could be attained:

1.1 Collection Management

AI can greatly improve data collection management, through automation.36 This means that data collection – through any type of device, such as Satellites, hidden microphones, algorithms scraping the internet… – could be scheduled, triggered automatically, or even forecast. AI could also decide what to collect, for instance knowing where a satellite should shoot pictures.

1.2 Signals Detection

Through the ubiquity of connected devices (computers, phones, IoT, CCTV) AI could help to surveil much more than any group of humans could attempt to.37 This allows for a much more nuanced monitoring of what is happening, and algorithms could sense anomalies in data as soon as they happen.38 As Katz formulates it: "As technical collection capabilities mature, data scientists and analysts can work to build, test, and hone AI models of likely enemy signatures, patterns, and activities and smartly search broad areas for priority targets."39 AI  could therefore enhance early warning systems.

1.3 Target Identification

Because of its ability to scan through vast amounts of data, AI has the potential to detect and pick new potential targets. This scenario could even be envisaged with human intelligence collection, through the use of electronic gear (such as Augmented Reality glasses). Machine Learning algorithms could thus recognize faces, buildings – any potential assets. 40

2. Processing and Exploitation

A growing problem regarding the processing and exploitation of data for intelligence is that "data in the IC are generated in too many diverse formats, in too many disconnected or inaccessible systems, without standardized structures and without overarching agreed-upon ontology".41 This explains why a lot of the collected data that might be valuable is never used – a lot of wasted potential. It also explains why this step of the intelligence cycle is the one that has gotten the most attention yet, with probably the biggest room for gains in efficiency. Attempts to implement AI aim to therefore improve processing in the following ways:

2.1 Triage

AI can help sort the Intelligence Community's data, by automatically discarding unusable bits of information and sorting it. As this is still often done manually, tremendous time gains can be expected.42

2.2 Task Automation

Machine Learning will help to automate tasks to process data, such as automatic tagging, labeling, recognizing patterns (e.g., recognizing a type of tank on a picture).43 This could even be envisaged for real time speech-to-text transcription, voice identification and language translation.44 This data could then even be clustered using neural networks to connect data in unforeseen ways. These explorative big data methods would then allow to infer conclusions.45

2.3 Visualization and Sensemaking

Machine Learning techniques could then, based on the clusters of data that have been inferred, automatically generate visualizations – network analyses, patterns, graphs…46

3. Analysis and Production

Using AI for the Analysis and Production stage will enable Intelligence officers to write reports faster.

3.1 Smart Search

Deep Learning is currently used to help understand search queries in search machines better and find better, more human, results, "(e.g., what is adversary X's strategy for Y?). Analysts could team with data scientists to tailor how data is tagged (e.g., words associated with "strategy") and how queries are sequenced to enable algorithms to learn and launch more complex or indirect searches".47

3.2 Data Visualization Tools

AI could help analysts designing their reports in compelling ways by enabling automatic graphs and other visualizations to be implemented for selected data (e.g., asking to input a timeline of a conflict – AI would then automatically craft one out of the information it can access).

3.3 Testing of Analytic Lines

While writing, one could envisage regression models that would confront the arguments that an analyst is advancing with their own estimations (e.g., when answering the question of a likelihood of a terrorist attack for a given place and time).  Therefore, analysts are constantly intellectually challenged to think about their arguments and avoid groupthink. Furthermore, "AI could surface anomalous, undervalued, and countervailing reports that analysts trusting a small, compartmented source base may have missed or discounted".49 This would enhance the fact-checking habits of intelligence officers even more.

3.4  Production Automation

Finally, AI could even automate the writing of bits and pieces of analysis, enabling analysts to concentrate on the pressing issues and highly complex questions. For instance, one could envisage a function to automatically input biographies in a piece, helping craft better reports faster.50

4. Dissemination

AI will also impact dissemination, by making finding and receiving intelligence reports much more user-friendly with the help of algorithms. One could think of almost Youtube-like features to facilitate the experience of finding what is needed.

4.1 Content Prioritization

Machine learning algorithms could help curate the reports that are currently most read or most important regarding the current agenda in order to help users to stay up to date and get the most relevant information at all times. This is something that is usually still done by hand, which is why precious time gains could be achieved.51

4.2 Content Personalization

AI could be set up to provide an individualized experienced when browsing Intelligence report, in order to "better time, tailor, and target products to diverse sets of consumers according to their unique intelligence needs".52 In this way, readbooks would be customized according to specific and personal interests.

4.3 Content Summarization

Machine Learning could then in turn make the vast number of reports more digestible, by summarizing them thanks to Machine learning and Natural Language Processing, combining them into smaller "readbooks".53

4.4 Content Enrichment

Finally, one could imagine AI also helping to augment and enrich existing reports, embedding additional, more recent data.54 In this way, live updates, for instance on case numbers during a pandemic, could update reports automatically.

5.  Planning and Direction

Planning and direction are the categories where AI will not have a particularly groundbreaking influence – for the moment. As a human in the loop is still very much desired when planning out the big picture, and AI still lacks multitasking and human thinking abilities, automatization is unlikely. However, it could still help to disseminate orders among personnel.

5.1 Granularity of Command Process

AI could be envisaged to automatically pick up and disseminate orders that have been issued and adapt them to the relevant situations and skillsets that are needed for the concerned personnel. For instance, an order to report on a protest in France might be automatically subdivided into more orders, one prompting HUMINT-agents to go investigate, the other prompting GEOINT-agents to take pictures from the sky.

Implications: A SWOT Analysis

Now that the different possible applications of AI in the Intelligence Cycle have been assessed, we can now look at the Strengths, Weaknesses, Opportunities and Threats that follow from this.

1. Strengths

Automation: Automation is and will be the greatest strength of using AI in the Intelligence Cycle, as up to 75% of the current work done by officers could be transferred to machines in the future.55 All cited examples involve AI automating tasks that used to be done by humans or could not even be done by them. Therefore, AI virtually is automation.

Velocity: Speed gains is another tremendous advantage of using AI for Intelligence. The time gains that can be made thanks to AI will empower countries that can harness it significantly more powerful relative to others.56

Scale: As we have seen, current data volumes are impossible to manage for humans working in intelligence. Automation and computing power therefore are a must in order to tackle the task of collecting intelligence in a meaningful way today, underlining the inevitability of AI in Intelligence.57

2. Weaknesses

 Focus on quantitative data over qualitative: Quantitative data is the focus of the technological revolution. AI is better suited to process quantitative data – numbers than qualitative data – ideas.  However, this can be a serious pitfall. As Krohley underlines it: Quantitative data explores the What, Where, Who. What we need however, are the Why and How of things.58 Furthermore, a lot of important answers lie within data-points that we do not yet possess or can't easily reach. AI is therefore incapable of retrieving some information (e.g., known by a handful of people, not published on the internet, not geographically locatable) that can be crucial – humans therefore stay relevant. "It is a question of adding an essential layer of depth and meaning to what has become a two-dimensional targeting process that is, in turn, driving an increasingly reductive and de-contextualized intelligence cycle."59

Increased costs: Because of the exponential expansion of data, ICs will have to allocate much more resources to server renting, which is costly.60

Inability of AI to innovate: AI is bad at innovation. First, "most AI systems today are trained to do one task, and to do so only under very specific circumstances. Unlike humans, they do not adapt well to new environments and new tasks." 61 This means that as soon as something unforeseen happens, humans will have to supervise the situation and assess if the AI will still output valid results. Second, this means that "humans are much better at what we haven't figured out yet":62  As long as AI is not fully sentient, a system solely based on AI will be fundamentally inapt to innovate

3. Opportunities

Reallocation of resources: Thanks to the strengths of AI, a reallocation of resources is possible, both financially and regarding time. Financially, less budget has to be devoted to menial tasks such as labeling, that used to be (and are often still) done by hand. The taxpayer's money is therefore more likely to contribute on impactful work done by intelligence officers. The argument is the same regarding time: Intelligence agencies can devote more of their time for strategic work, and less for preparatory details.

Improved real time analysis: Thanks to the gains in velocity, instantaneous data collection, processing and dissemination allows for improved real time analysis, leading to better on the fly decision-making.

New insights attainable: AI's explorative capabilities can harness new insights that were entirely out of reach before. Paired with other knowledge, such as social-behavioural sciences for instance, ML algorithms are able to deliver unique insights (e.g., how threat actors use technology).

4. Threats

We distinguish between external threats (pertaining to rivals) and internal threats (endogenous to an IC).

External Threats

Detection, Denial, Disruption, Deception: The ability to use AI for Intelligence becoming generalized. In fact, to prepare for possible disruptions, one has to envision a scenario where rival powers are at a similar capacity level, both for the offensive and for the defensive, and don't hesitate to employ their cyberweapons in unethical ways.

Detection & Denial: AI makes it therefore harder for intelligence agents to operate in foreign countries, both physically and digitally. This fragilizes the effectiveness of undercover agents (as AI surveillance, facial recognition etc. make it harder to impersonate someone else). This could create so-called denied areas for Allied Intelligence and pose the problem of a "vanishingly short shelf-life of secrets".63

Disruption: even simple, low-tech solutions are able to confuse and render AI useless, if employed in a clever way. Indeed, "simple obfuscation techniques such as strategic tape placement can fool AI", in the case of a tank and image recognition for example.64 Therefore, the threat that AI won't work because the enemy disrupts it is always a possibility.

Deception: more advanced and perverse than Disruption, Deception could be used by rivals to infect datasets used by AI and inject false data. "A plausible future scenario might involve an AI system that is charged with analyzing a particular question, such as whether an adversary is preparing for war. A second system, operated by the adversary, might purposefully inject data into the first system in order to impair its analysis."65 Machines will therefore spy on other machines. However, Deception can also work among humans, for instance by using so-called deep-fakes, AI-driven video & audio impersonations of people.66

Internal Threats

Bureaucratic Barriers: Bureaucratic Barriers could hinder the successful implementation of AI in the Intelligence Communities in two ways. First, bureaucratic culture could be a problem: Analysts having developed their techniques over decades might be unwilling to welcome radical change. "Deeply embedded in the analytic community are institutional, bureaucratic, and cultural preferences and bias toward the time-tested tradecraft and techniques they perceive to be the global gold standard."67 This might not even just be a problem of willingness: using AI requires a lot of technical skill. Therefore, senior analysts who are not proficient in digital tools but are essential to the functioning of intelligence agencies might pose a challenge to successful implementation.68 Second, experience has shown that, as the private sector is the driver of change, the IC has to rely on it to delegate a significant number of tasks related to AI. Public-Private Partnerships have therefore flourished.69 However, "obtaining […] data and sharing it is a challenge, especially for an organization that prefers to classify data and restrict access to it".70

Loss of Interoperability with Allies: Incompatible Digital Infrastructures could hamper cooperation among allies.71

Bad data/overcomplicated systems: AI offers many new possibilities but might also be overwhelming. This is for several reasons. First, because AI adapts to the data it analyzes, it is very untransparent. There is no guarantee that algorithms work properly or have not been spoiled by enemy data injections.72 Some results that seem dubious to personnel might therefore be unusable, because humans are unable to understand how the algorithms could have come up with such results.73 Second, even if the algorithm does work properly, it might be biased. Because humans set the priorities for the algorithms, they could unknowingly bias the ML-algorithms.74 Third, it seems ethically difficult to potentially delegate accountability to an algorithm. As a former intelligence official said about automation in an interview: "you could have one bad algorithm and you're at war." 75

Unethical usage of AI: Finally, as pictured by the Snowden leaks in 2013, AI usage might also drift off into a highly unethical practice, through a "near perfect surveillance".76 This might drive discussions about the legitimacy and boundaries of intelligence collection in democracies.

 5. SWOT Matrix

Outlook and Final Recommendations

AI will impact the Intelligence Cycle greatly, but quite differently when looking at the different steps of the Cycle. Some may remain primarily executed by humans (such as Planning and Direction or some aspects of Analysis and Exploitation), others might become even more automated, such as the Collection, Processing or Dissemination phases. The qualitative element of intelligence – especially in HUMINT – will not fade away. It would be foolish – even dangerous – to rely all too much on data alone. Current AI still has significant shortcomings. For instance, "to date, there is no evidence of a major improvement in [Electronic Warfare] system capability driven by the introduction of deep neural networks or deep reinforcement learning".77 Private-Public partnerships will also play an increasingly important role.78 Furthermore, "traditional boundaries that separate the HUMINT, SIGINT, and cyber disciplines"79  will blur. So will also the orderliness of the Intelligence Cycle, as AI will greatly speed up many aspects of intelligence. Therefore, AI is indeed a revolution in Intelligence affairs. However, it is not a wondrous problem solver. The future of intelligence will therefore most probably consist of a mix between AI tools and human decision-making, priority-setting and algorithmic design.80 Robots are far from ready to take over, but we need to know how to use them. Therefore, future work should rely on answering the question of "when, where, and how should a human be involved in each specific AI function"81, as suggested by Baker, in order to determine rules for who should be held accountable and when. Furthermore, the debate on the ethics of AI surveillance stays particularly relevant. In the meantime, countries should try to implement the following recommendations in order to successfully manage the challenges posed by AI:

1. Avoid over-reliance on AI – Countries should embrace technological change, but keep an eye on qualitative intelligence collection, as it often gives unique and crucial insights.

2. Look for better ways to implement public-private partnerships – As this is where the majority of innovations comes from, countries should capitalize on the private sector – tech firms, startups etc., while managing to keep their sensitive information for themselves. This is only possible through the elaboration of new working methods together with the private sector.

3. Establish technological synergies among Allies – Allies should cooperate where they can in order to strengthen interoperability. Allies could for instance work on common datasets and the unification of data classification methods for easier sharing.

4. Shift-focus to counter-deception – One of the most critical threats is the possibility of being deceived, by the injection of fake data or through techniques such as deep-fakes. Countries should therefore invest more resources into developing techniques to spot and stop deception attempts.

5. Rules are meant to be broken – but they frame the debate – Countries should push for more military diplomacy in order to set rules of engagement regarding AI and Intelligence with their partners and rivals. It would then be easier to hold rogue states accountable for misusing AI.


1 Brantly, A.F. (2018). "When Everything Becomes Intelligence: Machine Learning and the Connected World", Intelligence and National Security ,33(4), 7 June, p. 571.

2 Denécé, E. (2014). "The Revolution in Intelligence Affairs: 1989–2003", International Journal of Intelligence and CounterIntelligence ,27(1), March, pp. 27–41.

3 Brantly, 2018

4 Ibid., p. 570.

5 Gannon, J.C. (2008). "Managing Analysis in the Information Age", In Roger Z. George and James B. Bruce (eds.), Analyzing Intelligence: Origins, Obstacles, and Innovations, Georgetown University Press, pp. 213–25.

6 Being the different steps in the treatment of intelligence

7 Kojm, C. A. (2016). "Global Change and Megatrends - Implications for Intelligence and Its Oversight", In Zachary K. Goldman and Samuel J. Rascoff (eds.), Global Intelligence Oversight: Governing Security in the Twenty-First Century, Oxford: Oxford University Press, pp. 95–118.

8 Pendall, D. (2018). "Semper Optiones: 21st Century Intelligence", Air & Space Power Journal, Spring 2018, p. 27.

9 Baker, J.E. (2020). The Centaur's Dilemma: National Security Law for the Coming AI Revolution, Washington, DC: Brookings Institution Press, p. 1.

10 Direction du Renseignement Militaire (2019). "L'intelligence artificielle et ses enjeux pour la Défense", Revue Défense Nationale, p. 113.

11 Baker 2020, p.4

12 Direction du Renseignement Militaire 2019, p. 116.

13 Renault, V. (2020). "SWOT Analysis: Strengths, Weaknesses, Opportunities, and Threats", Community Tool Box - University of Kansas, Available here, (Accessed on: 9 December 2020).

14 Regens, J.L. (2019). "Augmenting Human Cognition to Enhance Strategic, Operational, and Tactical Intelligence", Intelligence and National Security, 34(5), 29 July, p. 682.

15 Hildt, E. (2019). "Artificial Intelligence: Does Consciousness Matter?", Frontiers in Psychology,10 (2 July), p.1.

16  Baker 2020, p. 13.

17 Ibid.

18 Ibid.

19 Ibid., p. 14.

20 Ibid., p. 14-16.

21 Warner, M. (2013), "The Past and Future of the Intelligence Cycle", In Mark Phythian (ed.), Understanding the Intelligence Cycle, Oxfordshire: Routledge, pp. 24–34.

22 Phythian, M. (2013). "Introduction - Beyond the Intelligence Cycle?", In Mark Phythian (ed.), Understanding the Intelligence Cycle, Oxfordshire: Routledge, p. 17.

23 Ibid., p. 19.

24 Ibid.

25 Warner 2013, p. 29-30.

26 Ibid., p. 30.

27 Warner 2013, p. 22.

28 Krepinevich, A. (1994). "Cavalry to Computer. The Pattern of Military Revolution", The National Interest.

29 Gray, C. S. (2002). "Strategy for Chaos: Revolutions in Military Affairs and the Evidence of History", Frank Cass.

30 Denécé 2014, p. 27.

31 Ibid.

32 Vinci, A. (2020). "The Coming Revolution in Intelligence Affairs", Foreign Affairs, 3 September, Available here, (Accessed on: 9 December 2020).

33 Ibid.

34 Gilli, A. (2019). "Preparing for 'NATO-Mation':" Nato Defense College Policy Brief, 4 (5), p.1

35 Walsh, P.F. (2021). Intelligence Leadership and Governance: Building Effective Intelligence Communities in the 21st Century, Milton: Taylor & Francis Group, p.119

36 Katz, B. (2020b). "The Collection Edge: Harnessing Emerging Technologies for Intelligence Collection", Center for Strategic and International Studies, 13 July, Available here, (Accessed on: 21 December 2020), p. 2.

37 Baker 2020, p. 31.

38 Ibid., p. 33.

39 Katz 2020b, p.2

40 Ibid., p. 3

41 Weinbaum, C. and Shanahan, J.N.T. (2018). "Intelligence in a Data-Driven Age", Rand Corporation, 13 July, Available here, (Accessed on: 9 December 2020), p. 5.

42 Katz, 2020b.

43 Walsh 2021, p.118.

44 Katz 2020b, p. 3.

45 Ibid.

46 Ibid.

47 Katz, B. (2020a). "The Analytic Edge: Leveraging Emerging Technologies to Transform Intelligence Analysis", Center for Strategic and International Studies, 9 October, Available here (Accessed on: 21 December 2020), p. 3

48 Walsh 2021, p. 123.

49 Katz 2020a, p.5.

50 Katz, B. (2020c).  "The Intelligence Edge: Opportunities and Challenges from Emerging Technologies for U.S. Intelligence", 17 April 17, Available here, (Accessed on: 21 December 2020), p.3.

51 Ibid, p. 3.

52 Ibid, p. 4.

53 Katz 2020a, p.3.

54 Ibid, p. 3.

55 McLaughlin, J. (2020). "Artificial Intelligence Will Put Spies Out of Work, Too", Foreign Policy (blog), Available here, (Accessed on: 9 December 2020).

56 Vinci, 2020.

57 Baker, 2020.

58 Krohley, N. (2017). "The Intelligence Cycle Is Broken. Here's How to Fix It", Modern War Institute, 24 October.

59 Ibid.

60 Leetaru, K. (2020). "The High Costs Of Hosting Science's Big Data: The Commercial Cloud To The Rescue?", Forbes, Available here, (Accessed on: 9 December 2020).

61 Maxwell, P. (2020). "Artificial Intelligence Is the Future of Warfare (Just Not in the Way You Think)." Modern War Institute, 20 April, Available at: /artificial-intelligence-future-warfare-just-not-way-think/,(Accessed on: 9 December 2020).

62 Solon, O. (2016). "These Engineers Are Developing Artificially Intelligent Hackers", The Guardian, 3 March, Available here, (Accessed on: 9 December 2020).

63 Katz 2020b, p. 4-6.

64 Maxwell, 2020.

65 Vinci, 2020.

66 Baker 2020, p. 34.

67 Katz 2020a, p. 7.

68 Ibid.

69 Katz 2020c, p. 3.

70 Maxwell, 2020.

71 Dufour, M. (2018). "Will Artificial Intelligence Challenge NATO Interoperability?", NATO Defense College.

72 Baker 2020, p. 47.

73 Direction du Renseignement Militaire 2019, p. 110.

74 Katz 2020, p. 6.

75 McLaughlin, 2020.

76 Baker 2020, p. 38.

77 Stefanick, T. (2019). "AI in the Aether: Military Information Conflict," The Brookings Institution, 4 December, p. 124.

78 Moutouh, H. and Poirot, J. (eds.) (2018). Dictionnaire du renseignement, Paris: Perrin, p. 474-475.

79 Katz 2020b, p. 4.

80 Regens 2019, p. 682.

81 Baker 2020, p. 29.

COVID-19 Capitalism in Raced Markets
An Interview with Robbie Shilliam and Ali Bhagat
SAIS Journal Editorial Staff
Fishermen for Foot Soldiers
Repercussions of the War for South China Sea Fisheries
Michael Hall
Global Value Chains, Volatility, and Openness
D’Aguanno et al.
Negotiation with Gangs
Adapting Pruitt to Explore State and Gang Negotiations
Alexandria Polk
European Academia is Ripe for Disruption
An Interview with Erik Jones
SAIS Journal Editorial Staff
The Evolution of TARGET2 Positions in the Eurozone
Moritz Osterhuber
Early Trends in Digital Climate Activism During the 2020 COVID-19 Lockdown
Sahar Priano
The Convergence of Political Values of Citizens Across EU Member States along EU Enlargement Rounds
Victor Vorsatz
The Longevity of Populism in Brazil
Matthew Hughes
The European Union's Strategic Autonomy in Geopolitical Digital Struggle
Giorgio Severi