INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

General (ret) Professor Teodor FRUNZETI, Ph.D^

(Academy of Romanian Scientists, 3 Ilfov, 050044, Bucharest, Romania,

email: secretariat@aosr.ro)

Captain Ilie IFTIME, Ph.D Candidate^

Abstract: One of the main current concerns of the global technology

industry and academic environment is the improvement of artificial intelligence

(AI) models and their integration into the multiple domains of society. In the case

of national and international security, the drive to assimilate AI is also evident

within intelligence structures, regardless of their form and specific profile

(services, agencies, directorates, etc.), owing to the need to ensure primacy in the

exploitation of new information by decision-makers, since „information is the

world’s most important and contested resource and timely data is the new oil”¹.

Thus, this article aims to identify a series of capabilities that AI can

provide within intelligence structures in an attempt to ensure informational

advantage.

Keywords: artificial intelligence, intelligence structures, intelligence cycle,

multiple sources, intelligence analyst, informational advantage.

DOI

10.56082/annalsarscimilit.2026.2.145

Introduction

The accelerated evolution of new technologies reaches its peak today

through artificial intelligence. Although the concept emerged in the

relatively distant past (for example, in John McCarthy and others, 1955,

paper A Proposal for the Dartmouth Summer Research Project on Artificial

Intelligence, where the hypothesis was advanced that a detailed description

of the processes of intelligence could provide the starting point for

simulating autonomous machine learning²), its rise in research and

development was marked by two long periods of stagnation, or „AI



Entitled Member of the Academy of Romanian Scientists, President of the Military

Sciences Section, Doctoral Supervisor at "CAROL I" National Defense University, email:

tfunzeti@gmail.com.



“Carol

I”

National

Defense

University,

Bucharest,

Romania,

email:

andreiaalbu1@gmail.com.

1

Rosenbach Eric, Mansted Katherin, The geopolitics of information, 2019, available at

https://www.belfercenter.org/sites/default/files/pantheon_files/files/publication/Rosenbach_

Commerce_Testimony_5March19_Final.pdf, p. 1, accessed on 13.06.2026.

2

*** A proposal for the darthmouth summer research project on artificial intelligence,

1955, available at https://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html,

accessed on june 13,2026.

145

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

winters” (the first from 1950 to 1980 due to the lack of computing power

required to train AI models, and the second from 1980 to 2010 due initially

to the absence of the internet and subsequently to insufficient digitalization,

as the very object of study was missing-digital archives/the concept of Big

Data).

At present, however, AI is becoming visible across all sectors of

security: political, military, economic, social, and environmental. The

appetite for its integration is increasing, from the micro level, where

analysts use it to perform routine tasks, to the macro level, where major

powers have established national strategies in this field and have even

classified it as a national priority-for example, China. This versatile

technology can relieve the human factor of a considerable amount of

repetitive and administrative tasks, support analytical processes, and provide

decision-making support.

Contemporary artificial intelligence is no longer merely at an

experimental stage; rather, it is successfully integrated into an increasing

number of intelligence systems, or is even updated and improved as new

extensions of its capabilities emerge. The advantages of using AI

technologies by intelligence structures are multiple: processing huge

volumes of data in a very short time, accelerating the intelligence cycle and

the OODA loop, enabling the real-time integration of multiple sources,

augmenting the capabilities of intelligence officers, increasing the level of

early warning by improving the ability to anticipate threats, preventing and

countering the rapid evolution of online disinformation and propaganda,

improving counterintelligence-specific activities, etc. In what follows, we

will argue why these capabilities are important for intelligence structures

and will illustrate them through successful models that have already been

integrated.

1. Processing much larger volumes of data in a much shorter

time (the main advantage of AI systems)

Whereas in the past intelligence services were confronted with a lack

of information, today, in addition to the classical process of collecting the

necessary data, they are confronted with an excess caused by the

proliferation of sources (military and commercial databases, sensors,

satellites, digital networks, social platforms, etc.) or by adversaries flooding

the information space with false, redundant, duplicate, and other types of

information. The analysis, filtering, and selection of the necessary

information and its transformation into relevant intelligence products has

become an increasingly difficult process to carry out solely by human

analysts, regardless of the size of the team. Thus, the integration of certain

artificial intelligence models, in order to take over various repetitive tasks,

collect data, and conduct its initial processing at a faster pace, has become

146

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

more than necessary. The role of the human analyst is thereby redefined

around tasks such as verification, interpretation, supplementation,

validation, and contextualization.

In this regard, intelligence structures have created and refined

various systems based on AI technologies that can operate with huge

quantities of data in a very short time. Examples include Maxar and Planet

Labs (in the geospatial domain, capable of analyzing terabytes of satellite

imagery daily and identifying movements, positions, structures, etc. at the

millisecond level)³, Charlotte AI (in the cybersecurity domain; prevention

and countering of attacks through the analysis of characteristics and their

comparison with global indicators of compromise, being 75% faster than a

specialized human team)⁴, Real Time – Regional Gateway (an AI-enhanced

system of the U.S. National Security Agency, used by intelligence services

to store, integrate, and process billions of dispersed metadata items-calls, e-

mails, digital ecosystems-with the purpose of faster target selection)⁵, etc.

The processing of large volumes of data in a very short time has

direct implications for the main activity of an intelligence structure, namely

the conduct of the intelligence cycle by increasing its efficiency. The use of

different specific AI systems can thus assist with or take over certain

analytical tasks, and „the stages can be carried out simultaneously, in real

time, and not only in a distinct and sequential manner”⁶. Furthermore, as a

subsequent effect of the foregoing, the time required to complete the OODA

loop (observe, orient, decide, act), developed by John Boyd, is also

shortened. Nevertheless, although the acceleration of warfare and the

compression of the decision-making process support the need to integrate

AI at this level⁷, in this case such integration is much more difficult to

achieve than within the intelligence cycle, because the loop requires a

3

*** Deep learning enables satellite-based monitoring of large populations of terrestrial

mammals across heterogeneous landscape, 2023, available at https://www.nature.com/-

articles/s41467-023-38901-y, accessed on june 7, 2026.

4

*** Accelerate security operations with generative AI, 2024, available at https://www.-

delltechnologies.com/asset/en-us/solutions/business-solutions/technical-

support/crowdstrike-charlotte-ai-datasheet.pdf, accessed on 09.06.2026.

⁵Lyons Jessica, 'Four horsemen of cyber' look back on 2008 DoD IT breach that led to US

Cyber

Command,

2024,

available

at

https://www.theregister.com/special-

features/2024/05/10/four-horsemen-of-cyber-recount-building-us-cyber-command/522406,

accessed on 08.06.2026.

⁶Dudley Craig, Lessons from SABLE SPEAR: The Application of an Artificial Intelligence

Methodology in the Business of Intelligence, published in Studies in Intelligence, vol. 65,

no. 1, p. 12, available at https://www.cia.gov/resources/csi/static/ArticleSableSpear-

ExperimentInAI.pdf, accessed on 06.06.2026.

7

Johnson James, Automating the OODA loop in the age of intelligent machines:

reaffirming the role of humans in command-and-control decision-making in the digital age,

2022, available at https://www.tandfonline.com/doi/full/10.1080/14702436.2022.2102486,

accessed on 12.06.2026.

147

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

stronger presence of the human factor, especially in the last three stages.

Strategic

interpretation,

contextualization,

the

understanding

of

ambiguities/uncertainties/non-linear situations and dynamic politico-

military arrangements, etc., remain tasks assigned strictly to the human

factor, at least for the time being.

2. More efficient integration of multiple sources

Some of the most important information is collected through open

sources (OSINT). This section comprises a huge and diverse volume of

data: press, mass media, databases from various fields, digital platforms,

social networks, satellite imagery, photos/videos provided by CCTV

cameras, etc. Analyzing these sources in real time and in a corroborated

manner can be an almost impossible task for a team of analysts. In this

respect, AI can process all of them in a very short time and, with the help of

predefined algorithms, can identify reference elements/patterns.

An example of integration within OSINT is the OSIRIS platform,

developed by the CIA’s Open Source Enterprise in 2024 for the U.S.

Intelligence Community. According to Jennifer Ewbank⁸, its main role is

not to provide intelligence products, but to help specialists more easily

understand immense volumes of open-source data (triage, translation,

transcription). With its help, analysis becomes more complex and faster,

thinking becomes more refined, assumptions are questioned, and the

analysis of alternative scenarios becomes more comprehensive.

The main models involved are Large Language Models (so that as

much data generated in various foreign languages as possible can be

incorporated), Machine Learning (to understand the context in which the

respective information was released, the correlations between it, and to

interpret user queries increasingly accurately), and generative AI (to provide

the requested product in the desired form).

As illustrated above, artificial intelligence is increasingly used in the

analysis of huge volumes of data collected through various methods used by

intelligence structures: OSINT, HUMINT, SIGINT, GEOINT, IMINT, etc.

Each domain has optimized this process through various AI models. The

element of novelty, however, lies in the attempts to generate increasingly

comprehensive intelligence products by harmonizing and unifying⁹all these

8

Ewbank Jennifer, The Role of Artificial Intelligence in the U.S. Intelligence Community:

Current Uses and Future Developments, 2024, available at https://www.-

aspeninstitute.org/wp-content/uploads/2024/10/Ewbank_Role-of-AI-in-USIC_Final.pdf,

accessed on 06.06.2026.

9

Brown Charlotte, Signal in Sync: OSINT, HUMINT, SIGINT, GEOINT in the modern

intelligence environment, 2025, available at https://mattermost.com/blog/signals-in-sync-

modern-intelligence-environment/, accessed on 02.06.2026.

148

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

data specific to each domain into comprehensive information capable of

highlighting overall, truthful, reliable, and real-time pictures.

Depending on the needs of the beneficiary intelligence structure,

various such sources may be combined and unitary architectures may be

formed. For example, in the case of maritime monitoring¹⁰, GEOINT-type

data (radar satellites that define the type of vessel, direction, and speed),

OSINT (data uploaded by vessels/operators to different platforms as a result

of traffic obligations), and SIGINT (intercepted radio communications) can

be operated by AI software simultaneously and in real time (examples:

Palantir Foundry/Gotham Maritime¹¹). Thus, “dark vessel” actions (the

intentional shutdown of a vessel’s own localization systems in order to

evade tracking by international authorities) can be countered. Another

example is the Ukrainian project Eyes on Russia¹², whose AI algorithms

integrate IMINT, OSINT, and GEOINT data in order to map the results of

the contemporary war in this region.

The final desideratum in integrating multiple data sources is the

creation of a mobile and robust AI architecture capable of incorporating all

these structures simultaneously in order to respond to queries from various

domains. This, however, is limited by the large resources required

(especially technological ones), by information systems that are not

interconnected across borders (the absence of membership in international

organizations/communities), and by actors’ internal/external policies

(“closed” regimes).

3. Augmenting intelligence officers

The evolution of the professional capabilities of intelligence officers

at present can also be achieved by assimilating new methods and means of

collecting and analyzing information with the help of artificial intelligence.

This direction of modernization for intelligence services should be a priority

in the current context of the rapid evolution of technologies. AI should not

be viewed as a super-technology intended to replace the human factor, but

as the most efficient tool currently available. The better the intelligence

analyst learns to work with it, the greater the increase in the quality of

intelligence products will be.

¹⁰Navulur Kumar, Data fusion and the future of geospatial intelligence, 2025, available at

https://synspective.com/blogs/2025/kumar_blog1/, accessed on 05.06.2026.

11

*** A Brief Analysis of Palantir Gotham: A Collaborative and Interactive Big Data

Visualization Analysis Software Based on Dynamic Ontology, 2024, accessed on

03.06.2026.

12

Kotaridis Ioannis, Integrating Earth observation IMINT with OSINT data to create

added-value multisource intelligence information: A case study of the Ukraine–Russia war,

2023, available at https://securityanddefence.pl/Integrating-Earth-observation-IMINT-with-

OSINT-data-to-create-added-value-multisource,170901,0,2.html, accessed on 09.06.2026.

149

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

The concept of human-AI teaming appears increasingly often in the

specialized literature. Perfect synergy between these two entities can be

achieved when the technological component reaches considerable

algorithmic performance, developed over time on the basis of prior training

(integration of multiple sources, understanding of reasoning, intentions, and

query formats, etc.). Effective collaboration within this binomial means

exceeding any individual human or technological capacity. Why is this

binomial necessary? Because the technological component cannot evolve on

its own to such a level, since to date the human ethical and moral

component has been attributed to it only in certain particular and limited

ways. This “last frontier”¹³is very difficult to overcome, because AI would

have to possess consciousness and a soul.

For the moment, the working spectrum of this hybrid team is limited

only by present technological developments. Nevertheless, integration has

occurred at all levels of an organization (planning, decision-making,

execution, etc.) and can be observed in simple work frameworks

(performing repetitive/administrative tasks) or complex ones (strategic

analysis, designing scenarios, predictions, and supporting decision-making).

If the artificial intelligence system can support the fulfillment of analytical

tasks and even decision-making in various ways, the human factor retains a

series of attributes that cannot be shared with the technological component:

responsibility, ethics, moral intuition, complete contextual judgment, the

legitimacy of action, supreme authority (AI can have only procedural

authority), meta-coordination (although AI excels at executing structured

tasks at scale, the adaptation and coordination of internal actions remain at a

low level), and critical decision-making¹⁴.

Some actors on the international stage are already implementing this

concept within their own intelligence structures. For example, the United

Kingdom’s GCHQ (Government Communications Headquarters) introduced

the Augmented Intelligence System¹⁵to support intelligence officers in

managing the growing volume and complexity of data and information so as

to improve the speed and quality of decisions. Similarly, in the case of

Chinese military intelligence, the Generative Artificial Intelligence Large

13

Siteanu Eugen, Frunzeti Teodor, Coșereanu Liviu, Artificial Intelligence – From

technological hopes to ethical concerns, 2023, available at https://www.researchgate.net/-

publication/367067594_Artificial_Intelligence_-

_From_Technological_Hopes_to_Ethical_Concerns, accessed on june 08,2026.

14

*** Toward a science of human–AI teaming for decision making: A complementarity

framework, 2026, published in Oxford Academic – Pnas Nexus, vol. 5, available at

https://academic.oup.com/pnasnexus/article/5/3/pgag030/8490283, accessed on 10.09.2026.

15

*** Pioneering a New National Security – The ethics of artificial Intelligence, 2021,

available at https://www.gchq.gov.uk/artificial-intelligence/index.html, accessed on june

04, 2026.

150

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

Model Online Monitoring and Early Warning System¹⁶, developed by

NORINCO Group (a state-owned company), may be mentioned. It is

capable of monitoring large quantities of online data, fusing them with other

internal data, analyzing them, and supporting the human factor in validating,

interpreting, and integrating the information necessary for the decision-

making process. It was developed by training the LLM with a huge amount

of data from the “-INT” spectrum (OSINT, GEOINT, HUMINT, etc.). This

model provides the analyst with a highly refined AI tool (multi-source

information and important historical data).

4. Increasing the level of early warning by improving the ability

to anticipate threats

Preventing strategic surprise is one of the objectives of any

intelligence structure. In this respect, artificial intelligence plays a

particularly important role in drafting situational awareness reports based on

continuous scanning of the security environment, identifying indicators

specific to threats, determining abnormal situations in the various sectors of

security, correlating certain data and information, etc.

The responsibility of intelligence analysts in this case is to evaluate

and validate the alerts produced by these AI systems, since errors may occur

as a result of insufficient historical data required for prior algorithm

training, the emergence of new situations/behaviors that may be

misinterpreted, or even adversary interference, either directly (cyberattacks

aimed at taking control of and manipulating the system) or indirectly

(flooding the online environment with erroneous information, carrying out

deceptive maneuvers/diversions in the physical domain) against them. In

this case, a situational awareness AI system can eliminate certain human

biases (emotional predispositions, support for certain hypotheses while

ignoring contrary evidence, failure to integrate all sources, fatigue, pressure,

etc.), but it can also create specific biases due to the circumstances

mentioned above.

An illustrative example of the above is the implementation of

artificial intelligence within the Zero Trust policy proposed by the Office of

the Director of National Intelligence (ODNI) in the United States. Within

the security architecture, AI is implemented across all processes:¹⁷joint

analysis of the global context, outlining the user’s historical behavior,

16

Haver Zoe, Artificial Eyes: Generative AI in China’s Military Intelligence, 2025,

available at https://assets.recordedfuture.com/insikt-report-pdfs/2025/ta-cn-2025-0617.pdf,

accessed on june 04, 2026.

17

Ajish Deepa, The significance of artificial intelligence in zero trust technologies: a

comprehensive review, published in Journal of Electrical System and Inf Technology, vol.

11, 2024, available at https://link.springer.com/article/10.1186/s43067-024-00155-z,

accessed on june 03, 2026.

151

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

assessing the operating parameters of the device, calculating the dynamic

score, providing decision-making support regarding the granting/blocking of

access, etc. Through all these measures and more, the speed of the system

increases, as does the efficiency of preventing and countering cyberattacks.

5. Preventing and countering the rapid evolution of online

disinformation and propaganda

The digital environment is becoming an increasingly important

battlefield for capturing the public’s attention and influencing its

perceptions, attitudes, and behaviors. Through the phenomenon of

radicalization, society becomes polarized, institutional trust declines, and

democratic processes are called into question. Certain artificial intelligence

models can be exploited in this respect by intelligence structures through

their integration into software for monitoring, detecting, and analyzing

online social platforms and other digital environments. Coordinated

disinformation campaigns conducted by various hostile actors can thus be

identified more easily because specific AI technologies can monitor a much

broader space and generate alerts in a much shorter time.

The imperative for intelligence structures to assimilate these

technologies stems from the fact that adversaries are already operating with

them for offensive purposes. Propaganda activities are carried out by these

actors on a very large scale, incorporating large language models¹⁸that

adapt content according to the target audience and its characteristics. Thus,

similar narratives have been observed spreading across the territories of

several states. In this respect, intelligence structures can monitor: the

coordination of certain news platforms, the propagation of narrative

patterns, the activation of certain account networks and the increase in their

activity (anti-EU/NATO, conspiratorial messages) before important events

(elections, summits, forums), etc.

An example of an artificial intelligence-based capability that can

identify, counter, and attribute coordinated disinformation and propaganda

is represented by the SemaFor (Semantic Forensics) technologies¹⁹

developed by the U.S. DARPA (Defense Advanced Research Projects

Agency) on the basis of an academic-industrial partnership. The public

presentation of the advantages of these technologies took place during the

international DEF CON 32 conference in 2024.

Moreover, these software systems may also incorporate a series of

automatic rapid response mechanisms that label the fake news in question

18

*** LLMs as information warriors? Auditing how LLM-powered chatbots tackle

disinformation about Russia's war in Ukraine, 2024, available at https://arxiv.-

org/pdf/2409.10697, accessed on june 02, 2026.

19

*** SemaFor program from DARPA, 2024, available at https://www.semaforprogram.-

com/analytic-catalog, accessed on june 11, 2026.

152

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

and immediately provide correct information from reliable sources, so that,

until the decision-maker adopts certain positions and takes measures, the

public is warned that the respective information is erroneous.

6. Improving activities specific to the counterintelligence domain

The phenomenon of digitalization entails, among other things, the

storage/migration of data and information into digital databases. Protecting

them becomes an increasingly difficult task as their volume and importance

grow, and as infrastructure expands and becomes interoperable with that of

other actors. In this respect, software based on artificial intelligence can

carry out preventive actions by monitoring and signaling unusual behaviors

of information networks, atypical migrations of files, the export of data in

small quantities but over the long term, suspicious access to databases, the

suspicious deletion of documents, increased interest in certain information,

amplified communications traffic between various entities, etc.; or it can

even undertake countermeasures, for example by blocking cyberattacks and

automating the response to the respective IT incident.

Constant supervision of personnel who have access to certain

databases and of their activity also remains a very important task. Espionage

and sabotage actions are based primarily on the recruitment of sources who

have access to these databases. Therefore, security checks carried out with

the help of artificial intelligence transform the process from a

periodic/occasional and reactive one into a continuous and predictive one.

Moreover, many more subjects can be analyzed simultaneously than a

classical security team could manage.

Thus, data relating to an individual’s profile, relational circle

(declared and undeclared), behavioral changes (significant fluctuations in

discourse, attitudes, emotions, and actions), and financial risk (a high risk

generated by contracted loans, various financial problems, etc., may

facilitate corruption and blackmail) can be identified and analyzed

automatically and in real time. In this respect, systems enhanced with AI

models have been developed, such as UAM (user activity monitoring),

responsible for recording, storing, and selecting data, and UEBA (user and

entity behavior analytics), used for the analytical component, which

compares a predefined “normal profile/history” model with each user’s

subsequent actions. The system identifies precisely those weak indicators

that, taken separately, mean nothing, but which, when correlated, signal

unusual behavior.

These types of systems are specific to each institution, being trained

on the basis of internal rules as well as the classical behavior of an

employee (the hours at which the platform is accessed, locations, temporal

indicators, the volume and flows of data being operated, usual queries, etc.).

In this case as well, the final decision regarding the labeling of a person as a

153

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

“security risk” still belongs to the human factor, since “false positive”

situations may arise (a personal event, such as a death in the family,

generating certain negative emotional and behavioral states, may be

misinterpreted by AI).

Nevertheless, entities in the intelligence field pay particular attention

to these types of systems, integrating and constantly modernizing them, as

in the case of the FBI’s contracting of the Insider Threat Management Suite

for UAM/UEBA Capabilities²⁰from the company Everfox in December

2025. The new system supports intelligence activity, the protection of

classified information, and counterintelligence.

Conclusions

The migration of data and information from physical media to digital

environments, as well as the accelerated increase in the complexity of the

relationships among them, have led to the need to process increasingly large

volumes of data in the shortest possible time. Within intelligence structures,

this capability is sometimes translated into the need to ensure a real-time

overall picture based on multiple sources. Informational advantage can be

ensured only through complete and timely information, and systems based

on artificial intelligence are becoming one of the most reliable tools

currently used in this respect.

At the same time, contemporary society is increasingly characterized

by a framework suffocated by the multitude of data surrounding us.

Identifying the necessary, truthful, and timely information is becoming an

increasingly difficult task in a space flooded with redundant information,

duplicates in different forms, fake news, and constant disinformation and

propaganda campaigns. Artificial intelligence can be used in this respect for

both offensive and defensive purposes. Intelligence structures must remain

aware of the adversary’s intentions and actions and develop/improve

capabilities to prevent and counter them.

The spectrum of AI integration is broad, ranging from the

automation of simple administrative/repetitive tasks in order to relieve the

human factor of simple duties and accelerate the workflow, to the

performance of extended analytical processes, and up to the provision of

support to the decision-maker through complex information (scenarios,

strategies, forecasts, etc.). Within this range, the augmentation of the

intelligence officer with these various technological abilities must be a

natural (learning/performance) and constant process.

In various domains, the human–AI working binomial appears as one

of the most efficient current models of symbiosis. While the technological

20

*** Contract for the acquisition of “Insider Threat Management Suite for UAM/UEBA

Capabilities” by the FBI, 2025, available at https://www.highergov.com/contract/-

15F06726P0000116/, accessed on june 10, 2026.

154

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

component represented by artificial intelligence contributes through the

unprecedented scale of data analysis and the speed with which this is

achieved, the human component completes the structure with exclusive

characteristics such as intuition, contextualization, ethical and moral

choices, and the capacity to react correctly when unpredictable situations

occur, etc.

Therefore, nowadays actors no longer question the efficiency of AI,

but rather its immediate yet responsible integration. Nevertheless, the

characteristic of responsibility outlines an unfair competitive framework

among actors, because the process of collecting, storing, and using data is

subject to different legislation when comparing a democratic state with an

authoritarian regime. This rivalry is also felt at the level of intelligence

structures, which are pressured to deliver quality intelligence products as

quickly as possible. Whoever first obtains and exploits certain information

can gain significant advantages in an increasingly dynamic, volatile, and

competitive security environment. Thus, today we are witnessing an

“algorithmic arms race,” which is no longer merely a concept situated at the

boundary between theory and practice, but a fact.

BIBLIOGRAPHY

*** LLMs as information warriors? Auditing how LLM-powered chatbots

tackle disinformation about Russia's war in Ukraine, 2024,

available at https://arxiv.org/pdf/2409.10697;

*** SemaFor program from DARPA, 2024, available at https://-

www.¬semaforprogram.-com/analytic-catalog;

*** Contract for the acquisition of “Insider Threat Management Suite for

UAM/UEBA Capabilities” by the FBI, 2025, available at

https://www.highergov.com/contract/15F06726P0000116/;

*** A proposal for the darthmouth summer research project on artificial

intelligence, 1955, available at https://www-formal.stanford.edu/-

jmc/history/dartmouth/dartmouth.html;

*** Deep learning enables satellite-based monitoring of large populations

of terrestrial mammals across heterogeneous landscape, 2023,

available at https://www.nature.com/articles/s41467-023-38901-y;

*** Accelerate security operations with generative AI, 2024, available at

https://www.delltechnologies.com/asset/en-us/solutions/business-

solutions/technical-support/crowdstrike-charlotte-ai-datasheet.pdf;

155

INTEGRATION OF ARTIFICIAL INTELLIGENCE WITHIN

INTELLIGENCE STRUCTURES

*** A Brief Analysis of Palantir Gotham: A Collaborative and Interactive

Big Data Visualization Analysis Software Based on Dynamic

Ontology, 2024;

*** Toward a science of human–AI teaming for decision making: A

complementarity framework, 2026, published in Oxford Academic

– Pnas Nexus, vol. 5, available at https://academic.oup.com/-

pnasnexus/article/5/3/pgag030/8490283;

*** Pioneering a New National Security – The ethics of artificial

Intelligence, 2021, available at https://www.gchq.gov.uk/artificial-

intelligence/index.html;

AJISH D., The significance of artificial intelligence in zero trust

technologies: a comprehensive review, published in Journal of

Electrical System and Inf Technology, vol. 11, 2024, available at

https://link.springer.com/article/10.1186/s43067-024-00155-z;

BROWN C., Signal in Sync: OSINT, HUMINT, SIGINT, GEOINT in the

modern

intelligence

environment,

2025,

available

at

https://mattermost.com/blog/signals-in-sync-modern-intelligence-

environment/;

DUDLEY C., Lessons from SABLE SPEAR: The Application of an

Artificial Intelligence Methodology in the Business of Intelligence,

published in Studies in Intelligence, vol. 65, no. 1, available at

https://-

www.cia.gov/resources/csi/static/ArticleSableSpearExperimentInA

I.pdf;

EWBANK J., The Role of Artificial Intelligence in the U.S. Intelligence

Community: Current Uses and Future Developments, 2024,

available

at

https://www.aspeninstitute.org/wp-

content/uploads/2024/10/Ewbank_Role-of-AI-in-USIC_Final.pdf;

HAVER Z., Artificial Eyes: Generative AI in China’s Military Intelligence,

2025, available at https://assets.recordedfuture.com/insikt-report-

pdfs/2025/ta-cn-2025-0617.pdf;

JOHNSON J., Automating the OODA loop in the age of intelligent

machines: reaffirming the role of humans in command-and-control

decision-making in the digital age, 2022, available at https://www.-

tandfonline.com/doi/full/10.1080/14702436.2022.2102486;

KOTARIDIS I., Integrating Earth observation IMINT with OSINT data to

create added-value multisource intelligence information: A case

study

of

the

Ukraine–Russia

war,

2023,

available

at

https://securityanddefence.pl/Integrating-Earth-observation-

IMINT-with-OSINT-data-to-create-added-value-

multisource,170901,0,2.html;

LYONS J., 'Four horsemen of cyber' look back on 2008 DoD IT breach that

led to US Cyber Command, 2024, available at https://www.-

156

General (ret) Professor Teodor FRUNZETI, Ph.D

Captain Ilie IFTIME, Ph.D Candidate

theregister.com/special-features/2024/05/10/four-horsemen-of-

cyber-recount-building-us-cyber-command/522406;

NAVULUR K., Data fusion and the future of geospatial intelligence, 2025,

available at https://synspective.com/blogs/2025/kumar_blog1/;

ROSENBACH E., MANSTED K., The geopolitics of information, 2019,

available

at

https://www.belfercenter.org/sites/default/files/-

pantheon_files/files/publication/Rosenbach_Commerce_-

Testimony_5March19_Final.pdf;

SITEANU E., FRUNZETI T., COȘEREANU L., Artificial Intelligence –

From technological hopes to ethical concerns, 2023, available at

https://www.researchgate.net/publication/367067594_Artificial_Int

elligence_-_From_Technological_Hopes_to_Ethical_Concerns;

157