Knowledge as power: the role of research in modern intelligence methods

Over the past ten years, France has sig­ni­fic­antly strengthened its capa­city to ana­lyse intel­li­gence, to open up cer­tain aspects of it to research, and to foster dia­logue with tech­no­logy stake­hold­ers. This pro­gress has also made it pos­sible to dis­tin­guish more clearly between three func­tions that ini­tial enthu­si­asm ten­ded to con­flate: pro­du­cing sci­entif­ic know­ledge on intel­li­gence; trans­form­ing an oper­a­tion­al need into a test­able prob­lem; and devel­op­ing, indus­tri­al­ising and main­tain­ing the res­ult­ing sys­tems. These func­tions must cooper­ate, but they do not draw on the same skills, the same time­frames or the same stand­ards of evidence.

Intel­li­gence Stud­ies occupy the first of these paths. Holeindre and Oudet point out that research­ers and prac­ti­tion­ers have ‘two dis­tinct voca­tions’ and must not be con­flated. Journ­als, sem­inars, accred­ited train­ing pro­grammes and net­works have gradu­ally estab­lished intel­li­gence as a cross-dis­cip­lin­ary field of sci­entif­ic study. Fol­low­ing a long peri­od of Brit­ish and Amer­ic­an influ­ence, French research now pos­sesses suf­fi­cient crit­ic­al mass to test the impor­ted cat­egor­ies against the back­drop of its own institutions.

The stra­tegic ques­tion is there­fore no longer wheth­er intel­li­gence stud­ies exist. It is to determ­ine how a sov­er­eign organ­isa­tion can exper­i­ment with sci­ence and industry without con­flat­ing know­ledge pro­duc­tion, oper­a­tion­al trans­form­a­tion and the devel­op­ment of capability.

The cre­ation of Inter­ax­ions in 2016 marks a telling start­ing point. Presen­ted as the DGSE’s think tank, the ini­ti­at­ive her­alds an inter­face with civil soci­ety as a whole, extend­ing bey­ond the aca­dem­ic world alone. The ambi­tion is bold: to facil­it­ate the exchange of prob­lems, know­ledge and per­cep­tions between a secret ser­vice and its wider environment.

From 2017 onwards, Inter­ax­ions’ pub­lic activ­it­ies gradu­ally focused on the pro­duc­tion and cir­cu­la­tion of know­ledge. The pub­lic­a­tions cata­logued in 2018 focused in par­tic­u­lar on the post-ISIS era, Afric­an Islam and the West­ern Balkans. In March 2019, a work­shop devoted to the manip­u­la­tion of inform­a­tion built on the ques­tions raised by the 2017 pres­id­en­tial elec­tion. Dis­in­form­a­tion and cog­nit­ive vul­ner­ab­il­it­ies thus entered a struc­tured research frame­work at a very early stage.

This tra­ject­ory reveals a clas­sic insti­tu­tion­al mech­an­ism. Among sev­er­al ini­tial object­ives, the one that already bene­fits from an estab­lished com­munity, val­id­a­tion rules and pub­lic­a­tion plat­forms tends to sta­bil­ise more quickly. The inter­face becomes primar­ily aca­dem­ic. How­ever, this suc­cess pro­duces a sep­ar­at­ing effect: sci­entif­ic study gains its autonomy, whilst meth­od­o­lo­gic­al trans­la­tion and oper­a­tion­al devel­op­ment fail to find such a clearly iden­ti­fied com­mon ground.

The Intel­li­gence Cam­pus, inaug­ur­ated in Creil in March 2017, har­bours a com­par­able ambi­tion in the tech­no­lo­gic­al sphere. Mil­it­ary per­son­nel, research­ers, industry rep­res­ent­at­ives and start-ups are expec­ted to con­trib­ute to data pro­cessing and innov­a­tion with­in the Mil­it­ary Intel­li­gence Dir­ect­or­ate. The ini­tial vis­ion is that of a con­tinuum: under­stand­ing, design­ing, exper­i­ment­ing and equipping.

In a strik­ing par­al­lel with Inter­ax­ions’ tra­ject­ory, roles are rap­idly becom­ing spe­cial­ised. The agree­ment with the CNRS facil­it­ates col­lab­or­a­tion between research­ers and oper­a­tion­al staff; indus­tri­al part­ner­ships cla­ri­fy require­ments and facil­it­ate pro­cure­ment. Appear­ing before the Nation­al Assembly’s Defence Com­mit­tee in March 2021, the Dir­ect­or of the DRM described the Cam­pus as a ‘unit for gath­er­ing know­ledge and mon­it­or­ing tech­no­lo­gic­al build­ing blocks’ and spe­cified that the DRM gen­er­ally acquires ‘soft­ware build­ing blocks’. The decree of 16^th Janu­ary 2025 main­tains the Cam​pus’s sup­port for innov­a­tion, but does not assign it a role com­bin­ing strategy, sci­entif­ic author­ity and indus­tri­al devel­op­ment. Its role is there­fore primar­ily that of a one-stop shop for iden­ti­fic­a­tion, com­mis­sion­ing and networking.

The part­ner­ship with Inria, in turn, sheds light on this spe­cial­isa­tion. Estab­lished in 2020, the Secur­ity and Defence Mis­sion organ­ises expert­ise, research and tech­no­logy trans­fer for the bene­fit of the Min­is­tries of the Armed Forces and the Interi­or. Developed in col­lab­or­a­tion with the DRM, IntelLab is presen­ted as a plat­form for sim­u­la­tion, train­ing and aware­ness-rais­ing that rep­lic­ates a cycle of multi-source stra­tegic inform­a­tion processing.

IntelLab thus helps to make intel­li­gence a share­able sci­entif­ic and tech­nic­al dis­cip­line. Scen­ari­os, ad hoc data, game design and debrief­ing make a gen­er­ic cycle vis­ible. In their 2023 art­icle, Crück and Asso­nion present the sim­il­ar­ity between sim­u­lated exper­i­ence and real-world activ­ity as a hypo­thes­is yet to be val­id­ated. The pub­lished work there­fore estab­lishes sig­ni­fic­ant achieve­ments in terms of accul­tur­a­tion, form­al­isa­tion and know­ledge. It estab­lishes neither a new oper­a­tion­al meth­od nor any meas­ur­able gains in terms of time, accur­acy or ana­lyt­ic­al robust­ness. The dis­tinc­tion is cru­cial: rais­ing aware­ness, demon­strat­ing and trans­form­ing a prac­tice are three dis­tinct levels.

The cre­ation of the Defence Innov­a­tion Agency in Septem­ber 2018 makes this dis­tinc­tion even clear­er. Stra­tegic research and intel­li­gence stud­ies have their own chan­nels; innov­a­tion pri­or­it­ises applic­a­tion, demon­strat­ors and scal­ing up. Between the two lies a trans­form­at­ive func­tion which Didi­er Baza­l­gette – draw­ing on nearly thirty years of coordin­a­tion between R&D, the DGA, the European Defence Agency and industry – pro­poses to organ­ise under the name ‘oper­a­tion­al sci­ences and development’.

This func­tion requires the accu­mu­la­tion of three forms of cap­it­al. The first is oper­a­tion­al, at a stra­tegic and meth­od­o­lo­gic­al level: under­stand­ing a ser​vice’s stra­tegic dir­ec­tion, its trade-offs, its doc­trines and its secur­ity con­straints. The second is sci­entif­ic: know­ing how to for­mu­late a test­able prob­lem, organ­ise a pro­gramme, super­vise research and assess the robust­ness of a res­ult. The third is indus­tri­al and entre­pren­eur­i­al: trans­form­ing evid­ence into a product that is main­tain­able, fund­able and integ­rable. Seni­or­ity alone does not guar­an­tee any of these trans­itions; hybrid­ity requires a sus­tained abil­ity to switch between dif­fer­ent lan­guages, meth­ods and cri­ter­ia for success.

Such a mod­el can pro­duce res­ults, but it can­not be scaled up when it relies on just a few individuals. 

Such a mod­el can pro­duce res­ults, but it can­not be scaled up when it relies on just a few indi­vidu­als. Research on bound­ary span­ners and role con­flicts shows that each domain demands its own proofs of com­pet­ence. Main­tain­ing oper­a­tion­al access, sci­entif­ic cred­ib­il­ity and indus­tri­al res­ults sim­ul­tan­eously requires not only con­tinu­ous legit­im­isa­tion, but also the ongo­ing main­ten­ance of three net­works, three lan­guages and three forms of author­ity. The organ­isa­tion thus places upon the hold­er of the role not only the cost of coordin­a­tion, but also the bur­den of main­tain­ing and renew­ing each of these forms of cap­it­al. One-off ini­ti­at­ives become pos­sible; their con­tinu­ity remains depend­ent on the avail­ab­il­ity of a single indi­vidu­al. The mod­el is there­fore excep­tion­al by design and can­not be rep­lic­ated by any organisation.

An addi­tion­al dif­fi­culty stems from admin­is­trat­ive cat­egor­isa­tions. As soon as an activ­ity involves a thes­is, a meth­od or a labor­at­ory, it tends to be equated with research. The cri­terion for suc­cess then shifts towards pub­lic­a­tion, leav­ing oper­a­tion­al devel­op­ment without a clearly des­ig­nated own­er. How­ever, assess­ing the value of a res­ult is not enough to trans­form it into capacity.

The Court of Aud­it­ors describes integ­ra­tion into rigid pro­gramme cycles as ‘com­plex and costly’; the Sen­ate expli­citly iden­ti­fies the dis­con­nect between exper­i­ment­a­tion and scal­ing up. Stand­ing between proof of concept and oper­a­tion­al imple­ment­a­tion are the budget, the con­tract, intel­lec­tu­al prop­erty, archi­tec­ture, secur­ity, doc­trine, train­ing and main­ten­ance. Research estab­lishes own­er­ship of the res­ult; it cre­ates neither the sys­tem into which it is integ­rated nor the sup­port for it. In 2024, the con­tinu­ation of this func­tion can there­fore no longer depend on a few excep­tion­ally hybrid indi­vidu­al career paths.

From 2023–2024 onwards, a dif­fer­ent archi­tec­ture becomes con­ceiv­able: no longer attempt­ing to insti­tu­tion­al­ise rare pro­files, but rather facil­it­at­ing the con­ver­gence of the three forms of cap­it­al – oper­a­tion­al, sci­entif­ic and indus­tri­al. In 2026, Baza­l­gette and Jan­in denounce ‘the illu­sion of pre­pared­ness’ arising from the con­fu­sion between nar­rat­ive frame­works and oper­a­tion­al com­pet­ence: foresight demon­strates through mean­ing; exper­i­ment­a­tion, through data. Time con­straints, endo­gen­ous feed­back and meas­ure­ment make sim­u­la­tion a space in which doc­trine, beha­viour, sci­entif­ic mod­els and tech­nic­al solu­tions can be tested together.

Sim­u­la­tion does not magic­ally recre­ate the con­tinuum envis­aged in 2016, nor does it replace the three forms of cap­it­al. It can, how­ever, provide them with a shared infra­struc­ture. The oper­at­or becomes a co-design­er: their tacit know­ledge is trans­lated into vari­ables, con­straints and per­form­ance cri­ter­ia. The research­er con­structs falsifi­able mod­els. The indus­tri­al part­ner devel­ops the engine, inter­faces, secur­ity, main­ten­ance and the con­di­tions for scal­ing up. A pro­tec­ted core can con­tain sens­it­ive data, rules and met­rics, whilst share­able sub-mod­els make it pos­sible to draw on civil­ian expert­ise without expos­ing the entire problem.

This archi­tec­ture also meets leg­al and organ­isa­tion­al require­ments. Nation­al defence secrecy neces­sit­ates com­part­ment­al­isa­tion and a ‘need-to-know’ approach; pub­lic pro­cure­ment reg­u­lates rela­tions with com­pan­ies. A mod­u­lar plat­form does not over­ride any of these rules: it can provide a tech­nic­al imple­ment­a­tion of them. It enables access levels to be segreg­ated, con­tri­bu­tions to be traced and dual-use com­pon­ents to be integ­rated with­in a con­trolled envir­on­ment. Civil soci­ety no longer merely provides ideas; it par­ti­cip­ates in a gov­erned devel­op­ment chain.

The assess­ment of the past dec­ade is instruct­ive because it clears up an ambi­gu­ity. France now pos­sesses a robust, legit­im­ate and sci­en­tific­ally autonom­ous field of Intel­li­gence Stud­ies. It also has aca­dem­ic and tech­no­lo­gic­al inter­faces that are use­ful to the intel­li­gence ser­vices. How­ever, suc­cess­ive attempts have failed to sta­bil­ise the trans­form­at­ive func­tion com­bin­ing intel­li­gence strategy, sci­entif­ic meth­od and indus­tri­al development.

For decision-makers, the les­son is simple. It does not lie in bet­ter select­ing, appoint­ing or pro­tect­ing a hand­ful of hybrid pro­files. Ten years of repeated fail­ures to sta­bil­ise this func­tion – in research, oper­a­tion­al organ­isa­tions and indus­tri­al sup­port – reveal the sys­tem­ic lim­it­a­tions of the mod­el. The com­bin­a­tion of oper­a­tion­al, sci­entif­ic and indus­tri­al cap­it­al is too com­plex to imple­ment to be rep­lic­ated at will; main­tain­ing all three sim­ul­tan­eously places the bur­den of safe­guard­ing access, skills, net­works and oper­a­tion­al con­tinu­ity on just a few indi­vidu­als. An organ­isa­tion based on such an excep­tion quickly becomes unsustainable.

The chal­lenge is there­fore to shift the unit of gov­ernance: no longer seek­ing to insti­tu­tion­al­ise the role that com­bines all three forms of cap­it­al, but to organ­ise their inter­con­nec­tion in a sus­tain­able manner.

The chal­lenge is there­fore to shift the unit of gov­ernance: no longer seek­ing to insti­tu­tion­al­ise the role that com­bines all three forms of cap­it­al, but to organ­ise their inter­con­nec­tion in a sus­tain­able man­ner. In 2026, sim­u­la­tion offers the most cred­ible frame­work for this trans­form­a­tion: not a fourth com­munity, but a shared infra­struc­ture in which the three strands remain autonom­ous whilst col­lect­ively pro­du­cing observ­able cap­ab­il­it­ies. After ten years of open­ness, the next step for­ward will not be a new circle. It will be the capa­city to exper­i­ment, meas­ure and industrialise.