Artificial Intelligence in F.R.S.-FNRS Grant Applications: Submission and Funding Patterns from 2016 to 2025

1 Introduction

The Fund for Scientific Research (F.R.S.–FNRS) is entrusted with a core mission: to advance fundamental scientific research driven by the initiatives and creativity of researchers themselves. As a major funding agency in the Wallonia-Brussels Federation (Fédération Wallonie-Bruxelles - FWB), the F.R.S.-FNRS supports the production of new knowledge across the full spectrum of scientific disciplines. Its funding model is rooted exclusively in scientific excellence and operates through various mechanisms, including the support of individual researchers, the funding of research teams, participation in international programmes, the provision of mobility grants, and the awarding of scientific prizes. The Fund also plays a key role in promoting European research and innovation programmes and in assisting researchers in their participation.

Each year, the F.R.S.-FNRS launches several calls for proposals covering all scientific domains - Exact and Natural Sciences (ENS), Human and Social Sciences (HSS), and Life and Health Sciences (LHS). Most of these calls follow a bottom-up approach, allowing researchers to submit proposals freely, without predefined thematic priorities. Complementary to this, associated specialised funds also support more targeted, top-down programmes addressing specific societal needs or areas with strong innovation potential.

The rapid rise of Artificial Intelligence (AI) worldwide has attracted significant attention across disciplines. In this context, the present analysis seeks to examine how AI-related research has evolved within F.R.S.-FNRS calls over the past decade. By analysing both submitted proposals and funded projects between 2016 and 2025, this study provides insights into the growing presence of AI in the research landscape and the corresponding investment made by one of Belgium’s principal funding agencies.

2 Methodology

2.1 Targeted data

This analysis examines funding applications submitted to the FNRS between 2016 and 2025, as summarized in Table 2.1. The calls considered are primarily recurring calls, with the exception of EOS and WISD, which were launched respectively twice and on a one-off basis.

The calls fall into two main categories: credits and projects, supporting scientific and technical staff, equipment, and operational costs; and grants and fellowships, providing personal support to researchers at various career stages (doctoral and postdoctoral researchers, or permanent positions). Renewal and promotion applications were excluded, as they relate to ongoing projects or mandates rather than new ones. The table lists each call along with the related funding instruments considered for the analysis. For further details on these instruments, please refer to the F.R.S.-FNRS website.

As mentioned earlier, the F.R.S.-FNRS funds research across all scientific domains:

• Exact and Natural Sciences (ENS), or in French, Sciences Exactes et Naturelles (SEN);
  
• Human and Social Sciences (HSS), or in French, Sciences Humaines et Sociales (SHS);
• Life and Health Sciences (LHS), or in French, Sciences de la Vie et de la Santé (SVS);
• Sustainability (SUST), which is not a scientific field in the strict sense; rather, at the F.R.S.-FNRS, it designates applications that address sustainability through an interdisciplinary approach and are regarded as forming a distinct domain in their own right.

While some calls specifically target certain domains, the main F.R.S.–FNRS calls (Credits and Projects, and Grants and Fellowships) cover the full range of these scientific areas.

Table 2.1: F.R.S.-FNRS calls and related instruments considered in the present analysis, conducted between 2016 and 2025

Funding category	Call	Call acronym	Related instruments	Scientific domains
Credits and projects	Credits and Projects call: range of funding schemes to provide scientific researchers who have a project of excellence with scientific and technical staff, equipment and means of operation	CREDIT	EQP, PDR, MIS, CDR, AMG-NEURO, AMG-PEDIATRICS	HSS, LHS, ENS, SUST
Credits and projects	The Excellence Of Science (EOS): programme covering all scientific fields and enabling collaboration between research teams from the Walloon and Flemish communities of Belgium, possibly complemented by federal and international research institutions (past call)	EOS	EOS, EOS-FULL	HSS, LHS, ENS
Grants and fellowships	FRESH: Human Sciences Research Fund for doctoral grants	FRESH	FRESH-B1	HSS
Grants and fellowships	FRIA: Fund for Research training in Industry and Agriculture for doctoral grants	FRIA	FRIA-B1	LHS, ENS
Credits and projects	Large Equipment (GEQ): programme designed to support the funding of major equipment as well as research infrastructures (past call, new version in 2024 entitled INFRA-GEQ)	GEQ	GEQ	HSS, LHS, ENS
Credits and projects	Infrastructure & Large Equipement (INFRA-GEQ): programme aimed at promoting and strengthening the state-of-the-art research infrastructure landscape in the Fédération Wallonie-Bruxelles by funding the acquisition of major equipment, the establishment and development of research infrastructures, as well as the upgrading of existing facilities	INFRA	INFRA-GEQ	HSS, LHS, ENS
Grants and fellowships	Grants and Fellowships call: funding instrument supporting individual researchers at different career stages (doctoral researchers, postdoctoral researchers, and permanent researchers)	MANDAT	CR, SD, ASP, CQ, CSD, SPD, MISU, SPD-REN2, SPD-REN3, VETE-CCD, CCL, CSD-REN2, CCL-REN, CSPD	HSS, LHS, ENS, SUST
Credits and projects	Thematic research project (PDR-THEMA): funding for research projects, either single- or multi-university, led by a principal investigator within a specific thematic area	PDR-THEMA	PDR-CARDIO	LHS
Credits and projects	Télévie: call aimed at supporting fundamental research through rigorously selected projects, with a view to advancing progress in the fight against leukemia and cancer in both children and adults	TELEVIE	TELEVIE, PDR-TLV, GRANT-REN-LUX, AMG-ONCO	LHS
Credits and projects	WEL-T: call supporting strategic, excellence-driven research in engineering sciences, chemistry and physics, with a view to valorising breakthrough innovations for industrial applications contributing to the sustainable transition	WEL-T	WEL-T ADV, WEL-T STG	ENS
Credits and projects	WELBIO: call supporting strategic, excellence-driven fundamental research in the life sciences, with a view to valorising its discoveries towards industrial applications in the field of health	WELBIO	AGR, CGR, SGR, ADV, STG	LHS
Credits and projects	FNRS WelCHANGE Programme: funding for projects in the human and social sciences with potential societal impacts	WELCHANGE	WELCHANGE	HSS
Credits and projects	WISD (Walloon Institute for Sustainable Development): axis of the Fund for Strategic Fundamental Research (FRFS) dedicated to sustainable development (only in 2016)	WISD	PDR-WISD	SUST

Depending on the scale of the call (number of applications and associated funding), one or several commissions/juries are responsible for the evaluation. More specifically, for the Credits and Projects as well as Grants and Fellowships calls, the candidates submit their application to one of the fourteen scientific commissions responsible for evaluation, this choice being made by the applicant on the basis of the scientific topic of the proposal. Given the relatively large number of submissions for these calls, this analysis will occasionally present figures disaggregated by scientific commission. The disciplinary scope of each commission is summarized in Table 2.2. It should be noted that the codes assigned to these commissions include a prefix corresponding to the French acronyms of the main scientific domains: SEN-* for Sciences Exactes et Naturelles, SHS-* for Sciences Humaines et Sociales, and SVS-* for Sciences de la Vie et de la Santé. Throughout this analysis, we retain indeed these commission codes as they are used by the F.R.S.-FNRS in its official documentation, including English-language documents.

Table of the F.R.S.-FNRS Scientific Commissions (Table 2.2)

Table 2.2: List of the fourteen scientific commissions of the F.R.S.–FNRS, along with a description of their disciplinary scope

Code	Description
SEN — Exact and Natural Sciences
SEN-1	Structure, electronic properties, fluids, nanosciences, biological physics; Analytical chemistry, chemical theory, physical chemistry/chemical physics; New materials and new synthetic approaches, structure-properties relations, solid state chemistry, molecular architecture, organic chemistry
SEN-2	All areas of mathematics, pure and applied, plus mathematical foundations of computer science, mathematical physics and statistics; Particle, nuclear, plasma, atomic, molecular, gas, and optical physics; Astro-physics/-chemistry/-biology, solar system, planetary systems, stellar, galactic and extragalactic astronomy, cosmology, space sciences, astronomical instrumentation and data
SEN-3	Informatics and information systems, computer science, scientific computing, intelligent systems; Electrical, electronic, communication, optical and systems engineering; Product and process design, chemical, civil, environmental, mechanical, vehicle engineering, energy processes and relevant computational methods; Advanced materials development: performance enhancement, modelling, large-scale preparation, modification, tailoring, optimisation, novel and combined use of materials, etc.
SEN-4	Physical geography, geology, geophysics, atmospheric sciences, oceanography, climatology, cryology, ecology, global environmental change, biogeochemical cycles, natural resources management; Ecology, biodiversity, environmental change, evolutionary biology, behavioural ecology, microbial ecology, marine biology, ecophysiology, theoretical developments and modelling; Biotechnology using all organisms, biotechnology for environment and food applications, applied plant and animal sciences, bioengineering and synthetic biology, biomass and biofuels, biohazards
SHS — Human and Social Sciences
SHS-1	Political sciences, international relations; Sociology, communication studies, science & technology studies; Social and cultural anthropology; Human and social geography, demography, health, sustainability science
SHS-2	Cognition; Psychology; Education sciences
SHS-3	Linguistics; Philosophy; Literature; Study of the arts, cultural studies
SHS-4	Historian approach of arts; History, archaeology; Religious studies
SHS-5	Economics; Finance, management; Law; Economic geography, demography, health, sustainability science, spatial analyses
SVS — Life and Health Sciences
SVS-1	Molecular biology, biochemistry, structural biology, molecular biophysics, synthetic and chemical biology, drug design, innovative methods and modelling; Genetics, epigenetics, genomics and other ’omics studies, bioinformatics, systems biology, genetic diseases, gene editing, innovative methods and modelling, ’omics for personalised medicine; Structure and function of the cell, cell-cell communication, embryogenesis, tissue differentiation, organogenesis, growth, development, evolution of development, organoids, stem cells, regeneration, therapeutic approaches
SVS-2	Organ and tissue physiology, comparative physiology, physiology of ageing, pathophysiology, interorgan and tissue communication, endocrinology, nutrition, metabolism, interaction with the microbiome, non-communicable diseases including cancer (and except disorders of the nervous system and immunity-related diseases); The immune system, related disorders and their mechanisms, biology of infectious agents and infection, biological basis of prevention and treatment of infectious diseases, innovative immunological tools and approaches, including therapies, veterinary medicine
SVS-3	Nervous system development, homeostasis and ageing, nervous system function and dysfunction, systems neuroscience and modelling, biological basis of cognitive processes and of behaviour, neurological and mental disorders
SVS-4	Medical technologies and tools for prevention, diagnosis and treatment of human diseases, therapeutic approaches and interventions, preventative medicine, epidemiology and public health, digital medicine, medical ethics; Pharmacy, pharmacology; Dentistry
SUST — Sustainability
SUST	The SUSTAINABILITY Commission of the F.R.S.-FNRS is committed to promoting excellent research on sustainability through interdisciplinarity. Sustainability is understood in a broad sense as encompassing the many challenges of sustaining human societies within planetary boundaries. Interdisciplinarity is understood as the articulation between disciplines usually addressed by different F.R.S.-FNRS thematic Scientific Commissions. The SUSTAINABILITY Commission itself is composed so as to correctly appreciate such an articulation. Projects submitted to the SUSTAINABILITY Commission should thus first aim at advancing sustainability, and second rely on at least two precisely articulated disciplines in doing so.

2.2 Remarks on the budget calculations

This analysis includes budget estimates. As indicated above, funding calls can be divided into two main categories:

• Credits and projects, for which the awarded amounts correspond to the budgets requested by the promoters at the time of application.

• Grants and fellowships, which correspond to salary-based funding and are subject to annual indexation. In this case, the calculated amount therefore corresponds to the sum of the annual (indexed) costs for the grant or fellowship holders over the theoretical duration of their mandate.

The costs presented in this analysis are theoretical estimates and should be interpreted as such:

• For credits and projects, it is possible that the full amount initially requested by the promoter was not ultimately spent.

• With regard to permanent researchers (CQ, Research Associates), we aggregated the costs associated with their positions up to and including 2025, without taking into account promotions that may have occurred during this period.

• In general, for grants and fellowships, we consider the theoretical duration for non-permanent positions, and the duration from the date of award up to 2025 for permanent positions. This methodology does not explicitly account for possible early terminations which may occur for various reasons (notably resignation), although such cases remain limited.

Disclaimer

The budget figures presented in this analysis should be understood as reasonable estimates derived from theoretical budget allocations and mandate durations, and not as expenditures.

2.3 Identification of AI-related applications

2.3.1 Guided by the use of descriptors and free keywords

The ERC/F.R.S.–FNRS classification is based on a set of descriptors used to characterise both the funding applications submitted to the Fund and the scientific expertise of the researchers in its expert database¹. Inspired by the ERC classification, it has been extended with additional FNRS-specific descriptors to better capture the specificities of research conducted in the Wallonia–Brussels Federation (FWB).

An updated version of this classification was adopted in 2023² to align with the latest ERC evolution and to reflect recent research trends in the FWB. This revision introduced, among others, new descriptor fields related to AI.

When submitting an application, applicants are required to select descriptors that best characterise their project. For the Grants and Fellowships and Credits and Projects calls, applicants must choose the scientific commission (see Table 2.2) - or, for the FRESH and FRIA calls, the jury - to which they wish to submit their proposal. At least two of the selected descriptors must correspond to the chosen commission/jury; otherwise, applicants must justify their choice. For the other calls, applicants are likewise required to select descriptors to characterise their application, even if the evaluation is conducted by a single commission in that case.

In principle, descriptor information can be used to identify proposals related to artificial intelligence (AI). However, this approach has several limitations. First, the descriptor classification is not always sufficiently granular, and descriptors explicitly dedicated to AI and machine learning were only introduced with the revised classification implemented in 2023. Moreover, given the requirement to select descriptors aligned with the chosen scientific commission, applicants may not systematically include AI-specific descriptors - associated with the SEN-3 commission - even when AI-related methods or components are involved. As a result, some AI-related proposals submitted to HSS or LHS commissions may not be captured through a descriptor-based approach alone. An alternative approach based on free-text keyword searches in proposal titles and abstracts is therefore necessary.

2.3.2 From two distinct perspectives

More broadly, identifying AI-related proposals proves to be challenging. Some projects consider AI as their primary research question. These projects are mainly situated within the field of technology, but they can also adopt philosophical, ethical, legal, epistemological, or any other critical or analytical perspective relevant to the human and social sciences. Such projects fall under a fundamental or exploratory approach, aiming to develop new knowledge around AI. Other projects focus on the application of existing AI methods to specific scientific problems, without introducing significant methodological innovation in the specific area of AI. These projects correspond to an applied or utilitarian approach, whose objective is to solve a concrete problem by using AI as a tool.

Although these two types of approaches differ in terms of their objectives and level of involvement in the AI field, they both fall within it. In the context of automated processing, it is therefore difficult to distinguish them with absolute certainty.

It may, however, be relevant to differentiate projects according to their disciplinary domain:

1. Projects with a technical or technological perspective, where AI is studied or developed within the framework of computer science or engineering, often with the goal of creating new methods, algorithms, or architectures.

2. All projects related to AI, across all scientific domains, where AI constitutes a central object of study or a core methodological component, including its development, application, or analysis in disciplines not only limited to computer science and engineering.

On this basis, the results presented in Section 3 will therefore be structured around these two categories, allowing for a differentiated examination of AI-related projects according to their disciplinary anchoring.

2.3.3 Based on the choice of a definition for AI

The identification work necessarily relies on a definition of artificial intelligence, which is itself not straightforward. There is indeed no consensus in the literature on a single definition of AI, and different levels of granularity can be adopted to delineate its scope. For the purposes of this study, a set of keywords was therefore selected as a proxy to represent this thematic area. This definition does not claim to be an absolute or authoritative reference; rather, it reflects a deliberate methodological choice, which could reasonably be subject to discussion.

A list of AI-related keywords (in their root form) was therefore defined to carry out this identification, both in English and French (see Table 2.3). Please note that the descriptors FNRS_58 and FNRS_59 were added to this list, as their labels correspond, in the new classifcation, to the respective domains of Machine Learning for Decision Making and Deep Learning, thus relevant for the identification task considered here.

Identification was carried out by searching through the titles, abstracts, free keywords, and descriptors associated with each application, in both French and English. Search keywords were therefore used in both languages to ensure maximal coverage of projects. As mentioned above, this identification exercise is carried out from two distinct perspectives, based on the disciplinary anchoring of AI projects.

To identify AI-related projects with a technical of technological perspective, the keywords listed in Table 2.3 were combined with the descriptor prefix “PE6_”, which corresponds - in both the old and new ERC/F.R.S.–FNRS classifications — to all the descriptors related to the fields of computer science, information systems, scientific computing and intelligent systems (including descriptors FNRS_58 and FNRS_59). Tables 2.4 and 2.5 present the descriptors related to the PE6 cluster, respectively in the former and new versions of the ERC/F.R.S.-FNRS classification. The request corresponding to the identification of the projects in the specific domains of computer science or engineering is presented below.

Table 2.3: Keywords used to identify AI-related applications, in their English and French versions when relevant. In some cases, only the English version is used, either because it is more commonly used in French contexts or because the English and French terms are identical.

Keyword (EN)	Keyword (FR)
Artificial intelligence	Intelligence artificielle
Artificial neural network	Réseau(x) de neurones artificiels
Chatbot	/
Computer vision	Vision par ordinateur
Convolutional neural network	Réseau(x) de neurones convolutif(s) / à convolution
Data mining	/
Deep learning	Apprentissage profond
Generative AI	IA générative
Large Language Model	/
Machine learning	Apprentissage automatique
Natural language processing	Traitement du langage naturel
Reinforcement learning	Apprentissage par renforcement
Supervised learning	Apprentissage supervisé
Unsupervised learning	Apprentissage non supervisé
Virtual assistant	/

Table of the F.R.S.-FNRS descriptors associated with the PE6 cluster (former classification 2016-2022) (Table 2.4)

Table 2.4: F.R.S.-FNRS descriptors associated with the PE6 cluster (classification considered before 2023)

Code	Descriptor (EN)	Descriptor (FR)
PE6_1	Computer architecture	Architecture informatique
PE6_2	Database management	Gestion de bases de données
PE6_3	Formal methods	Méthodes formelles
PE6_4	Graphics and image processing	Représentation graphique et traitement de l’image
PE6_5	Human computer interaction and interface	Interaction et interface homme-ordinateur
PE6_6	Informatics and information systems	Systèmes informatique et d’information
PE6_7	Theoretical computer science including quantum information	Informatique théorique y compris l’information quantique
PE6_8	Intelligent systems	Systèmes intelligents
PE6_9	Scientific computing	Informatique scientifique
PE6_10	Modelling tools	Outils de modélisation
PE6_11	Multimedia	Multimédia
PE6_12	Parallel and distributed computing	Informatique parallèle et distribuée
PE6_13	Speech recognition	Reconnaissance vocale
PE6_14	Systems and software	Systèmes et logiciels

Table of the F.R.S.-FNRS descriptors associated with the PE6 cluster (new classification 2023-2025) (Table 2.5)

Table 2.5: F.R.S.-FNRS descriptors associated with the PE6 cluster (new classification adopted in 2023)

Code	Descriptor (EN)	Descriptor (FR)
PE6_1	Computer architecture, embedded systems, operating systems	Architecture informatique, systèmes embarqués, systèmes d’exploitation
PE6_2	Distributed systems, parallel computing, sensor networks, cyber-physical systems	Systèmes distribués, informatique parallèle, réseaux de capteurs, systèmes cyber-physiques
PE6_3	Software engineering, programming languages and systems	Génie logiciel, langages et systèmes de programmation
PE6_4	Theoretical computer science, formal methods, automata	Informatique théorique, méthodes formelles, automates
PE6_5	Security, privacy, cryptology, quantum cryptography	Sécurité, vie privée, cryptologie, cryptographie quantique
PE6_6	Algorithms and complexity, distributed, parallel and network algorithms, algorithmic game theory	Algorithmes et complexité, algorithmes distribués, parallèles et de réseaux, théorie algorithmique des jeux
PE6_7	Artificial intelligence, intelligent systems, natural language processing	Intelligence artificielle, systèmes intelligents, traitement du langage naturel
PE6_8	Computer graphics, computer vision, multimedia, computer games	Infographie, vision par ordinateur, multimédia, jeux sur ordinateur
PE6_9	Human computer interaction and interface, visualisation	Interaction et interface homme-ordinateur, visualisation
PE6_10	Web and information systems, data management systems, information retrieval and digital libraries, data fusion	Web et systèmes d’information, systèmes de gestion de données, recherche d’informations, bibliothèques numériques, fusion de données
PE6_11	Machine learning, statistical data processing, signal processing (speech, image, video)	Apprentissage automatique, traitement statistique des données, traitement du signal (parole, images, vidéos)
PE6_12	Scientific computing, simulation and modelling tools	Informatique scientifique, simulation et outils de modélisation
PE6_13	Bioinformatics, bio-inspired computing, natural computing	Bioinformatique, informatique bio-inspirée, informatique naturelle
PE6_14	Quantum computing (formal methods, algorithms and other aspects of computer science)	Informatique quantique (méthodes formelles, algorithmes et autres aspects)
FNRS_58	Machine learning for decision making	Apprentissage automatique pour la prise de décision
FNRS_59	Deep learning	Apprentissage profond

Query considered to identify AI-related applications specifically associated with the PE6 cluster

(“artificial intelligence” AND PE6_) OR (“intelligence artificielle” AND PE6_) OR (“machine learning” AND PE6_) OR (“apprentissage automatique” AND PE6_) OR (“artificial neural network” AND PE6_) OR (“réseaux de neurones artificiels” AND PE6_) OR (“réseau de neurones artificiels” AND PE6_) OR (“natural language processing” AND PE6_) OR (“traitement du langage naturel” AND PE6_) OR (“computer vision” AND PE6_) OR (“vision par ordinateur” AND PE6_) OR (“reinforcement learning” AND PE6_) OR (“apprentissage par renforcement” AND PE6_) OR (“deep learning” AND PE6_) OR (“apprentissage profond” AND PE6_) OR (“chatbot” AND PE6_) OR (“virtual assistant” AND PE6_) OR (“generative AI” AND PE6_) OR (“IA générative” AND PE6_) OR (“large language model” AND PE6_) OR (“convolutional neural network” AND PE6_) OR (“réseaux de neurones convolutifs” AND PE6_) OR (“réseau de neurones convolutifs” AND PE6_) OR (“réseaux de neurones à convolution” AND PE6_) OR (“réseau de neurones à convolution” AND PE6_) OR (“supervised learning” AND PE6_) OR (“apprentissage supervisé” AND PE6_) OR (“unsupervised learning” AND PE6_) OR (“apprentissage non supervisé” AND PE6_) OR (“FNRS_58”) OR (“FNRS_59”)

The identification of AI-related projects across all scientific domains was only based on the keywords listed in Table 2.3 without the descriptor prefix “PE6_”, with the exception of the keyword reinforcement learning which on its own could be confused with a psychological concept. The query corresponding to the identification of the AI-related projects across all scientific domains is presented below.

Query considered to identify AI-related applications, across all disciplines

(“artificial intelligence”) OR (“intelligence artificielle”) OR (“machine learning”) OR (“apprentissage automatique”) OR (“artificial neural network”) OR (“réseaux de neurones artificiels”) OR (“réseau de neurones artificiels”) OR (“natural language processing”) OR (“traitement du langage naturel”) OR (“computer vision”) OR (“vision par ordinateur”) OR (“reinforcement learning” AND PE6_) OR (“apprentissage par renforcement” AND PE6_) OR (“deep learning”) OR (“apprentissage profond”) OR (“chatbot”) OR (“virtual assistant”) OR (“generative AI”) OR (“IA générative”) OR (“large language model”) OR (“convolutional neural network”) OR (“réseaux de neurones convolutifs”) OR (“réseau de neurones convolutifs”) OR (“réseaux de neurones à convolution”) OR (“réseau de neurones à convolution”) OR (“supervised learning”) OR (“apprentissage supervisé”) OR (“unsupervised learning”) OR (“apprentissage non supervisé”) OR (“FNRS_58”) OR (“FNRS_59”)

Disclaimer

The results of this study are inherently dependent on the definition of artificial intelligence adopted for the purpose of the analysis. This definition was adopted for methodological purposes and should not be regarded as an exhaustive or definitive representation of the field of artificial intelligence. As such, the results reflect this analytical choice and should be interpreted accordingly.

3 Analyses & Results

3.1 AI-related projects with a technical or technological perspective

3.1.1 Summary statistics

In total, among the 25617 applications considered in this analysis and meeting the inclusion criteria described in Section 2.1, 663 were identified as related to artificial intelligence with a technical or technological focus (see the associated query above).

Some key figures about these AI-related applications:

• They represent 2.6% of the proposals submitted to the F.R.S.-FNRS over the past decade (excluding, as mentioned in Section 2.1, renewal and promotion applications).

• Among these AI-related applications, 183 were awarded funding, representing 27.6% of the applications submitted to the F.R.S.-FNRS in this topic.

• They were submitted by 472 different researchers, encompassing all career stages (doctoral, post-doctoral, permanent).

• They were led by 238 different promoters (or principal investigators), i.e., researchers holding permanent academic or scientific position within a university of the Wallonia-Brussels Federation (FWB). Applicants to doctoral-level instruments must indeed be supported by a promoter, with whom they submit their application.

• Most of these applications were submitted under the Grants and Fellowships call (43.4%), which targets researchers at the doctoral and postdoctoral levels (see Figure 3.1). A substantial share also originates from the FRIA call (30.2%), intended for doctoral researchers working in the ENS and LHS domains. In addition, 21.0% of AI-related applications come from the Credits and Projects call, which is exclusively open to promoters holding permanent positions within universities of the Wallonia-Brussels Federation.

Figure 3.1: Percentage distribution of AI-related applications with a technical or technological perspective, by call, over the past decade 2016-2025 (n=663)

• Most of these applications fall within the domain of Exact and Natural Sciences (ENS) (84.2%). Figure 3.2 presents the distribution of AI-related applications across the different scientific domains. Unsurprisingly, most of the AI-related applications are linked to the SEN-3 commission (see Figure 3.3), whose disciplinary scope includes, among others, the fields of computer science and engineering. This outcome is expected, as the query used to identify AI-related applications with a technical and technological orientation was constructed using the descriptor fields of cluster PE6, which is directly associated with the disciplinary domain of the SEN-3 commission.

Figure 3.2: Percentage distribution of AI-related applications with a technical or technological perspective, by scientific domain, over the past decade 2016–2025 (n=663)

Figure 3.3: Percentage distribution of AI-related applications with a technical or technological perspective, over the past decade 2016–2025 (n=663)

3.1.2 Evolution over time

Figure 3.4 illustrates the evolution of the proportion of AI-related applications the last ten years, expressed as a percentage of the total number of applications received by the F.R.S.-FNRS. On average, 2.4% of the applications processed by the Fund fall within this domain. While this proportion may appear modest in absolute terms, it is far from negligible in practice, given that the Fund supports research across all scientific fields. Notably, this proportion reaches around 4% in 2023, which is consistent with the global turning point observed during the 2022–2023 period, marked by the public release and rapid dissemination of generative AI tools (notably ChatGPT in November 2022). However, caution is warranted when interpreting this increase, as 2023 also corresponds to the introduction of the new descriptor-based classification system. This structural change may itself have influenced the identification of AI-related applications and thus introduces a potential bias in the observations presented here.

It will be of interest to assess in the coming years how this trend will evolve, as there is not yet sufficient hindsight based on the observations from the subsequent years 2024 and 2025.

Figure 3.4: Annual percentage of AI-related applications with a technical or technological perspective among all the applications submitted to the F.R.S.-FNRS over the past decade 2016-2025

Figure 3.5 presents the evolution of the absolute number of AI-related applications, together with the number of applications that were granted funding.

Figure 3.5: Annual number of AI-related applications with a technical or technological perspective among all applications submitted to the F.R.S.-FNRS (purple bars), and annual number of those approved for funding (green curve) over the past decade 2016-2025

3.1.3 A brief overview of the descriptors

As mentioned earlier, this first part of the analysis is based on the identification of AI-related applications characterised by at least one descriptor related to the PE6 cluster. More generally, it is of interest to examine the extent to which the descriptors used to characterise these applications tend to co-occur in pairs, thereby highlighting the intensity of their interactions, and to observe whether descriptors originating from other descriptor groups emerge in this analysis.

We therefore examine descriptor co-occurrence frequencies on a pairwise basis. This analysis is restricted to applications submitted from 2023 onwards, corresponding to the year in which the new ERC/F.R.S.-FNRS classification was adopted. A total of 373 applications were therefore included in this identification process. As a result, the findings should be interpreted as indicative rather than exhaustive.

This approach made it possible to identify all occurrences of descriptor pairs across the applications considered. The maximum frequency observed for a given pair corresponds to 61 applications, which represents 16.4% of the analysed dataset, and concerns the FNRS_59 (Deep learning) - PE6_7 (Artificial intelligence, intelligent systems, natural language processing) pair.

Figure 3.6 schematically depicts the interactions between descriptors identified among the 373 analysed applications. For readability purposes, connections between descriptors are displayed in the graph only when their frequency of occurrence exceeds 10 applications, corresponding to 2.7% of the analysed applications. A table below the figure lists the descriptors along with their corresponding labels in English.

In particular, it is noteworthy that a core set of strong interactions emerges between the descriptors PE6_7, FNRS_58, FNRS_59 and PE6_11, all of which are related to machine learning and artificial intelligence. Some applications are also characterised by descriptors from the PE7 cluster, notably PE7_10, which is associated with robotics — an expected finding given the thematic focus of the present analysis. Furthermore, it is worth noting that a subset of applications is linked to the descriptor LS7_14 from the SVS-4 commission, corresponding to digital medicine and, more broadly, medical applications of artificial intelligence, as well as to the transversal descriptor IDR_2 related to big data. Finally, the results highlight the FNRS_55 descriptor, related to mathematical optimisation and operational research, which is typically associated with the SEN-2 commission. This observation can be readily explained by the strong connection between this discipline and decision-support methodologies.

Figure 3.6: Schematic representation of the descriptors reported in AI–related funding applications, with a technical or technological perspective (after the adoption of the new ERC/F.R.S.-FNRS classification in 2023, n=373). The figure displays the most frequent pairs of descriptors co-occurring within applications, with the thickness of the connecting curves indicating the frequency of occurrence of a given pair: the thicker the curve, the more frequently the pair of descriptors is observed

Code	Label
FNRS_55	Mathematical optimisation and operational research
FNRS_58	Machine learning for decision making
FNRS_59	Deep learning
IDR_2	Big data
LS7_14	Digital medicine, e-medicine, medical applications of artificial intelligence
PE1_18	Numerical analysis
PE6_10	Web and information systems, data management systems, information retrieval and digital libraries, data fusion
PE6_11	Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)
PE6_12	Scientific computing, simulation and modelling tools
PE6_2	Distributed systems, parallel computing, sensor networks, cyber-physical systems
PE6_3	Software engineering, programming languages and systems
PE6_6	Algorithms and complexity, distributed, parallel and network algorithms, algorithmic game theory
PE6_7	Artificial intelligence, intelligent systems, natural language processing
PE6_8	Computer graphics, computer vision, multimedia, computer games
PE6_9	Human computer interaction and interface, visualisation
PE7_10	Robotics
PE7_12	Electrical energy production, distribution, applications
PE7_3	Simulation engineering and modelling
PE7_7	Signal processing
PE8_4	Computational engineering

3.1.4 Budget

In total, 47.9 M€ were allocated to research in artificial intelligence for applications awarded between 2016 and 2025, corresponding to an average annual amount of 4.8 M€. The figure 3.7 illustrates the distribution of this funding across the different call types over the entire period (2016–2025).

Figure 3.7: Distribution of AI-related funding over the 2016–2025 period, by call type (over 183 funded applications)

The results show that the instruments of the Grant and Fellowship call (MANDAT) and the FRIA grants, which consist of salary costs, account for the largest proportion of the total budget (60.8%).

3.2 AI-related projects over all scientific domains

3.2.1 Summary statistics

In total, among the 25617 applications considered in this analysis and meeting the inclusion criteria described in Section 2.1, 1458 were identified as related to artificial intelligence over all scientific domains (see the associated query above).

Some key figures about these AI-related applications:

• They represent 5.7% of the proposals submitted to the F.R.S.-FNRS over the past decade (excluding, as mentioned in Section 2.1, renewal and promotion applications).

• Among these AI-related applications, 385 were awarded funding, representing 26.4% of the applications submitted to the F.R.S.-FNRS in this topic.

• They were submitted by 1011 different researchers, encompassing all career stages (doctoral, post-doctoral, permanent).

• They were led by 544 different promoters (or principal investigators), i.e., researchers holding permanent academic or scientific position within a university of the Wallonia-Brussels Federation (FWB). Applicants to doctoral-level instruments must indeed be supported by a promoter, with whom they submit their application.

• Most of these applications were submitted under the Grants and Fellowships call (45.5%), which targets researchers at the doctoral and postdoctoral levels (see Figure 3.8). A substantial share also originates from the FRIA call (21.9%), intended for doctoral researchers working in the ENS and LHS domains. In addition, 21.3% of AI-related applications come from the Credits and Projects call, which is exclusively open to promoters holding permanent positions within universities of the Wallonia-Brussels Federation.

Figure 3.8: Percentage distribution of AI-related applications over all scientific domains, by call, over the past decade 2016–2025 (n=1458)

• As shown in Figure 3.9, AI-related applications are unevenly distributed across scientific domains. Overall, the ENS domain accounts for just over half of all submissions (58.2%), followed by the LHS domain with 18.9%, and the HSS one, which represent 18.1% of the total.

Figure 3.9: Percentage distribution of AI-related applications, by scientific domain, over the past decade 2016–2025 (n=1458)

• A more detailed analysis of the commissions associated with the applications indicates that the SEN-3 commission and the FRIA are the most represented (see Figure 3.10), each accounting for around 22% of the AI-related submissions. It is worth noting that this finding was obtained even without linking research keywords (see Table 2.3) to the descriptors of the PE6 cluster. Nevertheless, this outcome is both logical and expected, given the nature of the topic under investigation, as the SEN-3 commission encompasses computer science and engineering and the FRIA call relates exclusively to applications within the ENS and LHS scientific domains. Other commissions also contribute to the evaluation of AI-related applications, albeit to a lesser extent, further illustrating the cross-disciplinary reach of artificial intelligence. The SEN-2 commission (mathematics, physics, and space sciences) covers 9.1% of the AI-related submitted applications. The SHS-5 (economics, finance, management, law, economic geography) and SVS-4 (clinical medicine, pharmaceutical sciences and dentistry) commissions follow, representing 5.9% and 5.0% of the identified applications, respectively.

Figure 3.10: Percentage distribution of AI-related applications over all scientific commissions, over the past decade 2016–2025 (n=1458)

3.2.2 Evolution over time

Figure 3.11 illustrates the evolution of the proportion of AI-related applications over the last ten years, expressed as a percentage of the total number of applications received by the F.R.S.-FNRS. On average, 5.3% of the applications processed by the Fund fall within this topic (across all scientific fields). Overall, the figure reveals a clear upward trend, with an approximate 2 percentage point increase observed in 2023, likely associated with the turning point marked by the emergence of GPT-based tools. However, as previously noted, the year 2023 also marks the implementation of the new ERC/F.R.S.-FNRS classification, which may have affected these results, with the introduction of descriptors more specifically dedicated to artificial intelligence.

The share of AI-related applications seems to continue rising in 2025 (10.3%), suggesting that the trend has not yet stabilised.

Figure 3.11: Annual percentage of AI-related applications across all scientific domains among all the applications submitted to the F.R.S.-FNRS over the past decade 2016-2025

Figure 3.12 presents the evolution of the absolute number of AI-related applications, together with the number of applications that were granted funding.

Figure 3.12: Annual number of AI-related applications over all scientific fields among all applications submitted to the F.R.S.-FNRS (purple bars), and annual number of those approved for funding (green curve)

3.2.3 A brief overview of the descriptors

As in the first part of this analysis, which focused on identifying AI projects with a technical and technological perspective (see more specifically Section 3.1.3), we propose to conduct a descriptor analysis of the identified AI-related projects across all scientific domains. Particular attention is paid to the most frequently observed pairs of descriptors among all applications. This approach is expected to reveal interactions and/or thematic groups involved in the context of AI-related research.

As explained before, this analysis is restricted to applications submitted from 2023 onwards, corresponding to the year in which the new ERC/F.R.S.-FNRS classification was adopted. As a result, a total of 818 applications were therefore included in this identification process. As a result, the finding should be interpreted as indicative rather than exhaustive.

The maximum frequency observed for a given pair corresponds to 61 applications, which represents 7.5% of the analysed dataset, and concerns the FNRS_59 (deep learning) - PE6_7 (artificial intelligence, intelligent systems, natural language processing) pair. This pair had already been identified as the most frequently observed among AI applications with a technical or technological perspective (see Section 3.1.3). Consequently, extending the analysis to AI-related applications across all scientific disciplines did not reveal any descriptor pair with a higher occurrence.

Figure 3.13 schematically depicts the interactions between descriptors identified among the 818 analysed applications. For readability purposes, connections between descriptors are displayed in the graph only when their frequency of occurrence exceeds 10 applications, corresponding to 1.2% of the analysed applications. A table below the figure lists the descriptors along with their corresponding labels in English.

Overall, the results highlight the existence of six distinct groups of applications.

• A cluster broadly similar to that identified in the analysis of descriptors focused on applications with a technical or technological perspective (see Figure 3.6) can be observed. This cluster is primarily characterised by PE6 descriptors, with strong interactions between PE6_7, PE6_11 and FNRS_59, which relate to machine learning and artificial intelligence. To a lesser extent, it also includes descriptors from PE1, PE7 and PE8 groups, drawn from other ENS commissions. The associated descriptors occasionally co-occur with the transversal field IDR_2 (related to big data). In addition, this cluster includes combinations involving LS7-type descriptors from the SVS-4 commission, and notably a strong interaction between descriptors LS7_14 (Digital medicine, e-medicine, and medical applications of artificial intelligence) and LS7_2 (Medical technologies and tools—including genetic tools and biomarkers - for the prevention, diagnosis, monitoring and treatment of diseases).

• A notable co-occurrence is observed between the HSS descriptors FNRS_40 (Legal theory, sociology of law, legal history, and philosophy of law), FNRS_43 (Public law), FNRS-47 (European law) and FNRS_49 (New technologies and artificial intelligence law). Their combination likely relate to legal issues/questions arising in the context of artificial intelligence.

• The association of the descriptors LS5_5 (Neural networks and plasticity), LS5_16 (Systems and computational neuroscience) and LS5_18 (Innovative methods and tools for neuroscience) appears to represent research focused on innovative AI-based techniques for neuroscience.

• Linguistics-related research is also identified among the results, through the co-occurrence of the SH4_9 (Theoretical linguistics; computational linguistics) and FNRS_16 (Corpus linguistics, lexicography and terminology) descriptors.

• Finally, two additional clusters can be identified in the domain of the SEN-2 commission. The first is related to particle physics, through the association of the PE2_2 (Phenomenology of fundamental interactions) and PE2_3 (Experimental particle physics with accelerators) descriptors. The second corresponds to astronomy research, which also emerges in the results via the association of the PE9_12 (High-energy and particle astronomy) and PE9_13 (Astronomical instrumentation and data, e.g. telescopes, detectors, techniques, archives, analyses) descriptors.

Figure 3.13: Schematic representation of the descriptors reported in artificial intelligence–related funding applications (after the adoption of the new ERC/F.R.S.-FNRS classificatioon in 2023, n=818) The figure displays the most frequent pairs of descriptors co-occurring within applications, with the thickness of the connecting curves indicating the frequency of occurrence of a given pair: the thicker the curve, the more frequently the pair of descriptors is observed

Code	Label
FNRS_16	Corpus linguistics, lexicography and terminology
FNRS_40	Legal theory, sociology of law, legal history, philosophy of law
FNRS_43	Public law (constitutional, human and administrative law)
FNRS_47	European law
FNRS_49	New technologies and artificial intelligence law
FNRS_55	Mathematical optimisation and operational research
FNRS_58	Machine learning for decision making
FNRS_59	Deep learning
FNRS_76	Translational research
IDR_2	Big data
LS2_11	Bioinformatics and computational biology
LS4_12	Cancer
LS5_11	Neurological and neurodegenerative disorders
LS5_16	Systems and computational neuroscience
LS5_18	Innovative methods and tools for neuroscience
LS5_5	Neural networks and plasticity
LS7_1	Medical imaging for prevention, diagnosis and monitoring of diseases
LS7_10	Preventative and prognostic medicine
LS7_14	Digital medicine, e-medicine, medical applications of artificial intelligence
LS7_2	Medical technologies and tools (including genetic tools and biomarkers) for prevention, diagnosis, monitoring and treatment of diseases
PE1_17	Mathematical aspects of computer science
PE1_18	Numerical analysis
PE2_2	Phenomenology of fundamental interactions
PE2_3	Experimental particle physics with accelerators
PE6_10	Web and information systems, data management systems, information retrieval and digital libraries, data fusion
PE6_11	Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)
PE6_12	Scientific computing, simulation and modelling tools
PE6_2	Distributed systems, parallel computing, sensor networks, cyber-physical systems
PE6_3	Software engineering, programming languages and systems
PE6_6	Algorithms and complexity, distributed, parallel and network algorithms, algorithmic game theory
PE6_7	Artificial intelligence, intelligent systems, natural language processing
PE6_8	Computer graphics, computer vision, multimedia, computer games
PE6_9	Human computer interaction and interface, visualisation
PE7_10	Robotics
PE7_12	Electrical energy production, distribution, applications
PE7_3	Simulation engineering and modelling
PE7_7	Signal processing
PE8_4	Computational engineering
PE8_5	Fluid mechanics
PE9_12	High-energy and particle astronomy
PE9_13	Astronomical instrumentation and data, e.g. telescopes, detectors, techniques, archives, analyses
SH4_9	Theoretical linguistics; computational linguistics

3.2.4 Budget

In total, 85.8 M€ were allocated to research in artificial intelligence for applications awarded between 2016 and 2025, corresponding to an average annual amount of 8.6 M€. The figure 3.14 illustrates the distribution of this funding across the different call types over the entire period (2016–2025).

Figure 3.14: Distribution of AI-related funding over the 2016–2025 period, by call type (over 385 funded applications)

The results show that the instruments of the Grant and Fellowship call (MANDAT) and the FRIA grants, which consist of salary costs, account for the largest proportion of the total budget expenditure (55.0%). However, budget requests submitted under the Credits and Projects call also represent a non-negligible proportion of overall costs (27.0%).

4 Discussion and Conclusion

This analysis examined both researchers’ interest in artificial intelligence and the level of F.R.S.-FNRS investment in this thematic area. It was structured around two complementary perspectives: first, applications in which AI is addressed from a technical and technological standpoint, and second, applications involving AI across all scientific domains.

Placing these two approaches side by side and comparing the resulting indicators highlights several key findings.

• Applications related to AI in a broad sense represent a little more twice the volume of applications focusing on AI from a strictly technical or technological perspective. A similar ratio is observed for the number of applicants and promoters involved, reflecting the broad interest in artificial intelligence across scientific disciplines, beyond projects focusing exclusively on its technical or technological development.

• Across both analytical scopes, the largest share of applications is submitted under the Grants and Fellowships call, which primarily targets early-career and non-permanent researchers. Notably, the proportions are very similar in both cases, with 43.4% of AI-related projects with a technical or technological perspective and 45.5% of AI-related projects across all scientific disciplines submitted through this call. A substantial proportion of applications also originates from the FRIA call, accounting for 30.2% and 21.9% of submissions, respectively.

• When considering AI-related applications across all domains, around 60% of the applications fall within the Exact and Natural Sciences, which is consistent with the technical and computational nature of the field. Beyond this domain, the remaining applications are distributed in a broadly balanced manner between the Life and Health Sciences and the Human and Social Sciences, which account for 18.9% and 18.1% of AI-related applications respectively over the last decade. This distribution highlights the transversal and increasingly interdisciplinary nature of AI-related research.

• Applications with a technical or technological focus on AI increased in relative terms, reaching a peak of around 4% in 2023. When considering AI-related applications across all scientific domains, a comparable surge is also observed in 2023, with these applications accounting for 8.2% of total submissions. At this stage, however, it remains premature to draw firm conclusions regarding future trends. It will nevertheless be of interest to monitor the evolution of these figures in the coming years. Moreover, the increase observed in 2023 may partly reflect the rapid emergence and dissemination of generative AI tools, but it may also be influenced by the transition to the new ERC/F.R.S.-FNRS classification system. As a result, disentangling the respective effects of these factors remains challenging, and the observed increase should therefore be interpreted with caution.

• The analysis of descriptor pair co-occurrences among applications submitted from 2023 onwards - following the introduction of the new ERC/F.R.S.-FNRS classification - made it possible to identify a range of thematic areas in which artificial intelligence is applied or investigated. Beyond the fields traditionally associated with computer science and engineering, the results also highlight AI-related applications in law, astronomy, particle physics, linguistics, neuroscience, and the health domain more broadly.

• Finally, from a budgetary standpoint, F.R.S.-FNRS funding allocated to AI-related research over the last decade amounts to 47.9 M€ for projects with a technical or technological focus, and to approximately twice this amount (85.8 M€) when AI-related projects across all scientific domains are taken into account. These figures should be interpreted in relation to the average annual FNRS budget, which is estimated at around 230 M€. Over a ten-year period, this corresponds roughly to one fifth of a single annual FNRS budget devoted to technically focused AI research, and around 40% of an annual budget when considering AI-related research across all domains. These levels of investment are therefore far from negligible and illustrate the strategic importance of AI within F.R.S.-FNRS funded research.

It is important to acknowledge that this study is subject to several limitations. First, the results depend on the definition of artificial intelligence adopted in this analysis. In the absence of a formal consensus within the literature, this definition is necessarily based on a methodological choice, which may partially bias the findings presented here. In addition, the change in the descriptor classification implemented in 2023 may have influenced the observed results, particularly by affecting the identification and aggregation of AI-related applications. Finally, it should be emphasised that the budgetary figures reported in this study are estimates and do not correspond to exact expenditure amounts.

Analyses & studies - FNRS

This report was written by the F.R.S.-FNRS Analysis, evaluation and Foresight unit.
The F.R.S.-FNRS Board of Trustees meeting of April 9, 2026 took note of these analyses.

The F.R.S.-FNRS, via its Analysis, evaluation and Foresight unit, carries out a certain number of analyses (mainly statistical ones): call analyses - following the closure of any call and making it possible to monitor its proper functioning; survey reports on former fellows and permanent researchers of the F.R.S.-FNRS and its associated Funds; survey reports on experts who have taken part in the various evaluation missions of the Fund; bibliometrics; report on the state of gender equality, etc.

For more information, please consult: FNRS Analyses & Studies page.

Contact data :
Sarah Itani (sarah.itani@frs-fnrs.be)
Analysis, Evaluation and Foresight
F.R.S.-FNRS
Rue d’Egmont 5, 1000 Bruxelles
Belgique

To cite this document: Fonds de la Recherche Scientifique-FNRS (2026). Artificial Intelligence in F.R.S.-FNRS Grant Applications: Submission and Funding Patterns from 2016 to 2025.

---

1. An analysis of the F.R.S.-FNRS expert database, with facts and figures, is available online (see link).↩︎

2. The list of descriptors associated with this classification is available online (see link).↩︎