Cupać, Jelena; Sienknech , Mi ja
A icle — Published Ve sion
Regula e agains he machine: how he EU mi iga es AI
ha m o democ acy
Democ a iza ion
P o ided in Coope a ion wi h:
WZB Be lin Social Science Cen e
Sugges ed Ci a ion: Cupać, Jelena; Sienknech , Mi ja (2024) : Regula e agains he machine: how he
EU mi iga es AI ha m o democ acy, Democ a iza ion, ISSN 1743-890X, Taylo & F ancis, London,
Vol. 31, Iss. 5, pp. 1067-1090,
h ps://doi.o g/10.1080/13510347.2024.2353706
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/308004
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
RESEARCH ARTICLE
Regula e agains he machine: how he EU mi iga es AI
ha m o democ acy
Jelena Cupać
a
and Mi ja Sienknech
b
a
WZB Be lin Social Science Cen e , Be lin, Ge many;
b
Eu opean Uni e si y Viad ina F ank u ,
F ank u , Ge many
ABSTRACT
Democ acies a e unde a ack om a ious sides. In ecen yea s AI-powe ed
echniques such as p ofiling, a ge ing, elec ion manipula ion, and massi e
disin o ma ion campaigns ia social bo s and oll a ms challenge he e y
ounda ions o democ a ic sys ems. Agains his backg ound, demands o
egula ing AI ha e go en loude . In his pape , we ocus on he Eu opean Union
(EU) as he ac o ha has gone he u hes in e ms o egula ing AI. We he e o e
ask: Wha kind o ins umen s does he EU en ision in hei binding and non-
binding documen s o p e en AI ha m o democ acy? And wha c i ique can be
o mula ed ega ding hese ins umen s? To add ess hese ques ions, he a icle
makes wo con ibu ions. Fi s , by building on a sys ema ic unde s anding o
delibe a i e democ acy, we in oduce he dis inc ion be ween wo ypes o ha m
ha can a ise om he widesp ead use o AI: igh s-based ha m and sys emic ha m.
Second, by analysing a numbe o EU documen s, including he GDPR, he AI Ac ,
he TTAP, and he DSA, we a gue ha he EU en isions ou p ima y ins umen s
o sa egua ding democ acy om he ha m ul use o AI: p ohibi ion, anspa ency,
isk managemen , and digi al educa ion. While hese ins umen s p o ide a
ela i ely high le el o p o ec ion o igh s-based AI ha m, he e is s ill ample space
o hese echnologies o p oduce sys emic ha m o democ acy.
ARTICLE HISTORY Recei ed 31 Oc obe 2022; Accep ed 7 May 2024
KEYWORDS A ificial In elligence (AI); democ acy; ha m; Eu opean Union (EU); AI Ac
In oduc ion
Conce ns abou a ificial in elligence (AI) nega i ely impac ing democ acy a e no
new. The Camb idge Analy ica scandal has al eady e ealed how p ac ices such as psy-
chog aphic p ofiling and a ge ed poli ical ad e ising could be used o elec ion
manipula ion. Howe e , ecen ad ancemen s in AI, no ably he eme gence o La ge
Language Models (LLMs) such as Cha GPT, along wi h models capable o gene a ing
images, audio, and ideo con en , ha e escala ed hese conce ns o a new le el. As hese
echnologies con inue o sp ead and cap u e public a en ion, we a e heading owa d
© 2024 The Au ho (s). Published by In o ma UK Limi ed, ading as Taylo & F ancis G oup
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.
o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k
is p ope ly ci ed. The e ms on which his a icle has been published allow he pos ing o he Accep ed Manusc ip in a eposi o y
by he au ho (s) o wi h hei consen .
CONTACT Jelena Cupać[email p o ec ed]
DEMOCRATIZATION
2024, VOL. 31, NO. 5, 1067–1090
h ps://doi.o g/10.1080/13510347.2024.2353706
he fi s wa e o elec ions in which hey will be widely deployed. I is expec ed ha
qui e a ew poli icians will use syn he ic con en in hei campaigns, whe he o
launch a acks agains hei opponen s h ough manipula ed ideos and image y o
o educe campaign cos s. Agains his backd op, we ocus on he Eu opean Union
(EU) and ask wha measu es i en isions o mi iga e he ha m AI poses o democ a ic
p ocess.
The EU’s ela ionship wi h democ acy is no pe ec . Since i s ounding, i has been
accused o s uc u al democ a ic defici and, mo e ecen ly, i has had o g apple wi h a
populis wa e causing democ a ic backsliding in se e al o i s membe s a es. S ill,
he e is no denying ha he EU is inex icably linked o democ acy. Along wi h
human digni y, eedom, equali y, he ule o law, and human igh s, he Lisbon
T ea y lis s democ acy as one o he EU’s co e alues. An impo an implica ion o
his alue o ien a ion is ha when he EU en u es in o a new egula o y e i o y,
i is expec ed o show conce n o democ acy and de ise mechanisms o i s p o ec ion,
especially i he e is ample e idence ha a gi en domain migh suffe significan demo-
c a ic e osion.
As he global ace owa ds AI in ensifies, we a e wi nessing a pa allel ace owa ds
AI egula ion, wi h he EU eme ging as a clea on unne .
1
Al hough lagging behind
he Uni ed S a es and China in AI de elopmen , he EU is ac i ely s i ing o es ablish
i sel as a global s anda d-se e in digi al echnologies. I has al eady achie ed his goal
h ough i s flagship Gene al Da a P o ec ion Regula ion (GDPR), wi h a ious
coun ies ou side Eu ope adop ing simila da a p o ec ion ules. The EU now seeks
o eplica e his impac wi h he Digi al Se ices Ac (DSA), aimed a bols e ing
online sa e y by ackling ha m ul con en , and mo e significan ly, wi h he A ificial
In elligence Ac (AIA), a comp ehensi e egula o y amewo k o AI g ounded in a
isk-based app oach.
2
Despi e hese documen s’obliga ion o add ess ha m o democ acy, a sys ema ic
o e iew o he measu es hey es ablish o his pu pose is s ill missing. As a esul ,
we ha e a limi ed unde s anding o he se o ac ions he EU and i s membe s a es
ha e a hei disposal o uphold democ a ic p ocesses in he AI age. The apid p og ess
o AI echnologies unde sco es he impo ance o such unde s anding, no leas o
iden i y which a eas o democ acy a e adequa ely p o ec ed and which equi e
u he sa egua ding.
To add ess his gap, we s a by iden i ying he ypes o ha m a widesp ead use o AI
can cause o democ acy. Based on a sys ema ic unde s anding o delibe a i e democ-
acy, we dis inguish be ween igh s-based ha m and sys emic ha m. The o me pe -
ains o using echnology o limi people’s pa icipa ion in he democ a ic p ocess,
while he la e e e s o b oade socie al and poli ical ac o s ha impede democ a ic
delibe a ion, such as agmen a ion, pola iza ion, dis us , and poli ical apa hy. In
making his dis inc ion, we join a g owing numbe o schola s who call o a
b oade app oach o AI ha m, one ha goes beyond indi idual and human igh s con-
ce ns and pays a en ion o he wide socie al impac .
3
Based on his amewo k, we ha e selec ed ou legally binding EU documen s o
ou analysis: he GDPR, DSA, AIA, and he P oposal o a Regula ion on he T ans-
pa ency and Ta ge ing o Poli ical Ad e ising (TTPA), alongside a ange o non-
binding documen s (see Appendix, Table 1). Th ough close examina ion o hese
documen s, we ha e iden ified ou key ins umen s he EU p oposes o p o ec
democ acy om AI ha m: p ohibi ion, anspa ency, isk managemen , and digi al
1068 J. CUPAĆAND M. SIENKNECHT
educa ion. P ohibi ion en ails he ou igh banning o specific da a and AI p ac ices;
anspa ency manda es he disclosu e o in o ma ion conce ning he de elopmen
and deploymen o AI sys ems; isk managemen in ol es assessing he isk o an
AI sys em be o e deploymen and moni o ing he isk a e he sys em is in use;
and digi al educa ion aims o aise public awa eness o AI isks and empowe indi-
iduals o use digi al echnologies esponsibly. Al hough hese ins umen s affo d a
ela i ely obus le el o p o ec ion agains igh s-based ha m, ou co e obse a ion
is ha he widesp ead use o AI s ill has significan po en ial o inflic sys emic ha m
o democ acy in he EU.
Be o e p oceeding, i is impo an o no e his s udy’s limi a ions. Ou p ima y
goal was o ca ego ize AI ha m o democ acy and ake a bi d’s eye iew o he
EU’s egula o y and policy esponses. Ideally, we would ha e ma ched each ype
o ha m wi h specific p o ec i e measu es and p o ided a de ailed e alua ion.
Howe e , he in e wined na u e o igh s-based and sys emic ha m, which is also
mi o ed in egula o y and policy ins umen s, makes such an app oach challenging.
Thus, we ha e op ed o a b oade app oach, which can se e as a guide o u u e
in-dep h analysis.
The a icle p oceeds in h ee s eps. Fi s , we define democ acy and AI, and discuss
igh s-based and sys emic ha m in mo e de ail. Second, we ou line ou analy ical s eps:
selec ing EU documen s and iden i ying ins umen s o shield agains AI ha m. Thi d,
we p esen and c i ically assess hese ins umen s. In conclusion, apa om summa -
izing ou findings, we del e in o he challenges and possibili ies o egula ing agains
AI ha m, he eby unde sco ing a significan a enue o u u e esea ch.
A ificial in elligence and democ acy: defini ions
To explo e he po en ial nega i e impac o AI on democ acy, we mus fi s define
bo h concep s. Howe e , defining AI and democ acy is no o iously difficul . AI can
e e o a b oad ange o cons an ly e ol ing echnologies and applica ions ha
mimic human cogni ion, such as lea ning, p oblem-sol ing, and decision-making.
Likewise, democ acy emains elusi e gi en he di e se ways democ a ic p inciples
mani es ac oss a ious socie ies and cul u es. Recognizing hese challenges, we
ha e op ed o a p agma ic app oach, seeking defini ions ha nei he impede
ho ough analysis no es ic b oade insigh s in o he EU’s AI egula ion o
democ acy p o ec ion.
Defining a ificial in elligence and i s applica ion in poli ics
AI can be defined as “Sys ems ha display in elligen beha iou by analysing hei
en i onmen and aking ac ions –wi h some deg ee o au onomy – o achie e
specific goals.”
4
AI sys ems can make au onomous decisions in a ious fields o
human ac i i y and gene a e o iginal con en such as ex , pic u es, and ideos.
Unlike con en ional digi al echnologies based on fixed algo i hmic pa e ns, AI
sys ems ely on neu al ne wo ks –machine lea ning algo i hms modelled a e he
human b ain’s s uc u e and unc ion – o lea n, adap , and e ol e h ough exposu e
o la ge amoun s o da a. In his way, AI sys ems can analyse complex pa e ns, make
in o med decisions, and c ea e con en ha exceeds he limi s o hei ini ial
p og amming.
DEMOCRATIZATION 1069
Cu en ly, nume ous AI echnologies and hei applica ions a e seen as po en ial
h ea s o democ acies. Some a e en i ely AI-based, while o he s ely on adi ional
digi al me hods. Ye , when augmen ed by AI, hese con en ional me hods can be sig-
nifican ly enhanced, heigh ening hei po en ial isk. B oadly speaking, hese echnol-
ogies can be spli in o wo ca ego ies along hei in ended unc ion: hose cen ed on
su eillance, such as da a collec ion and p ofiling, and hose gea ed owa ds manipu-
la ion, encompassing a ge ing, social bo s, and deep akes.
5
P ofiling consis s o collec ing and analysing da a abou people in o de o classi y
hem based on specific cha ac e is ics.
6
In he elec o al p ocess, demog aphic p ofiling
can be used o ca ego ize o e s by a ibu es like age and educa ion, whe eas psycho-
me ic p ofiling p obes deepe in o pe sonali y ai s, such as being a gumen a i e o
complian . Ye , p ofiling is no solely in o ma ional; i ypically suppo s a ge ing,a
echnique in ended o influence people’s a i udes and poli ical s ances.
7
While
mic o- a ge ing elies on his da a o iden i y a specific audience segmen , hype - a -
ge ing akes i a s ep u he by elying on e en mo e de ailed in o ma ion such as
loca ion acking and social media ac i i y.
Social bo s a e au oma ed p og ammes ha emula e human ac ions on social
media.
8
Augmen ed by AI models such as GPT and Midjou ney, hey could gene a e
and sp ead as amoun s o ex , images, ideos, and links. Wi h he abili y o ampli y
pa icula na a i es, sway discussions, and se e as a ehicle o sp eading disin o ma-
ion, social bo s pose a significan h ea o he in eg i y o online in o ma ion. The
same applies o deep akes, mul imedia coun e ei s ha swap one pe son’s image o
ideo likeness o ano he using ad anced algo i hms ha analyse hei acial cues.
9
The esul is a ake ye highly ealis ic piece o media.
I is c ucial o highligh ha we a e also en e ing an e a o apid de elopmen in AI-
powe ed neu o echnology, biome ics, and subliminal messaging, all expec ed o
u he exace ba e issues ela ed o p ofiling and manipula ion.
10
Technologies
anging om eal- ime biome ic iden ifica ion sys ems o subliminal messaging
such as AI-d i en “da k pa e ns”a e all poised o ampli y he accu acy and sub le y
o p ofiling and manipula ion, gene a ing a hos o new and o en un o eseeable chal-
lenges.
11
Las ly, al hough no ye a eali y, he e is significan discussion abou he
de elopmen o gene al-pu pose AI –sys ems capable o lea ning and execu ing a
b oad spec um o asks ins ead o being confined o specific oles o a eas. How
such AI will be in eg a ed in o poli ical sys ems is difficul o p edic , bu i s impac
on dis up ing adi ional democ a ic p ocesses and ci izen engagemen will be p o-
ound, so a ca e ul conside a ion o i s implica ions o go e nance, democ acy,
e hics, and socie y a la ge will be pa amoun in shaping i s ole and impac on poli ical
p ocesses.
Defini ion o democ acy: a sys emic delibe a i e app oach
To define democ acy, we ake a cue om he wo k o Spence McKay and Ch is
Teno e, who se led on he concep o sys emic delibe a i e democ acy in hei analysis
o how disin o ma ion affec s he democ a ic p ocess.
12
The ad an age o his
app oach o democ acy is ha a he han ocusing on he in icacies o mul iple
democ a ic a enas and p ocesses, i emphasizes delibe a i e unc ions ha uni e
hese a enas and p ocesses in o a sys emic whole. As a esul , a key ea u e o his
app oach o democ acy is iden i ying no ma i e goods and unc ions ha sus ain
1070 J. CUPAĆAND M. SIENKNECHT
meaning ul delibe a ion ac oss he sys em and which, i jeopa dized, can h ea en
democ acy in i s en i e y.
Diffe en au ho s emphasize diffe en no ma i e goods and unc ions essen ial o
genuine democ a ic delibe a ion.
13
In his s udy, we d aw on he wo k o Jane Mans-
b idge and he colleagues, whose app oach seems pa icula ly pe inen o he in o -
ma ion age. They iden i y h ee unc ions o delibe a i e democ acy: democ a ic,
epis emic, and e hical.
14
The democ a ic unc ion ocuses on he equal pa icipa ion
o ci izens, s essing he ac i e inclusion o di e se pe spec i es and he p e en ion
o unjus exclusion. The epis emic unc ion pe ains o he quali y o he in o ma ion
en i onmen in which delibe a ion akes place, especially he quali y o he in o -
ma ion and logics influencing p e e ences and opinions. Finally, he e hical unc ion
emphasizes mu ual espec in poli ical discussions, wi h he emphasis on pa icipan s
ecognizing each o he ’s iewpoin s as alid and dese ing o conside a ion.
These h ee unc ions sugges ha he bed ock o a h i ing democ acy lies in uphold-
ing pa icipa ion igh s, whe he o indi idualso g oups, and in he sys emic condi ions
gua an eeing his pa icipa ion. The sys emic condi ions de i e om he in eg i y o he
in o ma ion ecosys em and he quali y o in e ac ions among ci izens. Agains his back-
g ound, we assume ha AI can unde mine democ acy a i s co e bo h by cu ailing ci i-
zens’pa icipa ion igh s and by in oducing dis up ions a he sys emic le el. In o he
wo ds, we diffe en ia e be ween wo ypes o AI induced ha m o democ acy: igh s-
based ha m and sys emic ha m. In so doing, we align wi h he bu geoning li e a u e
seeking o map, sys ema ize, and axonomize AI ha m ac oss a ious domains.
15
A he same ime, we align ou sel es wi h a g owing numbe o sociologis s, legal
schola s, and policy expe s who, d awing on ounda ional insigh s om Zemiology
and Science and Technology S udies, a gue ha social ha m, whe he s emming om
AI o o he sou ces, should no be iewed in p edominan ly indi idualized and
human igh s e ms.
16
Ins ead, special a en ion should also be paid o socie al
ha m, including ha m o sys ems and s uc u es upholding socie ies.
17
We con ibu e
o his s and o li e a u e by dis inguishing be ween igh s-based and sys emic ha m
wi hin he specific con ex o democ acy, ins ead o add essing i h ough he lens o
a socie y as a whole.
18
This is an analy ical mo e ha we hope p o ides a amewo k
om which AI ha m o democ acy can be s udied in a b oad ye mo e ocused
manne .
AI ha m o democ acy: wo ypes
Righ s-based ha m o democ acy
In he con ex o his a icle, igh s-based ha m o democ acy e e s o using AI ech-
nologies o iola e indi idual, collec i e, and g oup igh s essen ial o ensu ing he
meaning ul pa icipa ion o di e se oices and pe spec i es in a democ a ic
socie y.
19
Indi idual igh s and eedoms, such as p i acy, eedom om disc imi-
na ion, and eedom o hough and exp ession, a e widely ecognized as essen ial
o democ a ic pa icipa ion. Collec i e igh s belong o g oups based on sha ed iden-
i y, such as language o cul u e, and a e i al o p o ec ing mino i y igh s and ensu -
ing hei ep esen a ion. Finally, g oup igh s apply o any g oup o indi iduals and
encompass such igh s as o ming associa ions and o ganiza ions and engaging in col-
lec i e ba gaining and ad ocacy.
DEMOCRATIZATION 1071
Based on he ex an li e a u e and eal-wo ld cases, we can disce n se e al scena ios o
digi al echnologies powe ed by AI causing significan ha m o hese igh s. One such
ins ance is he po en ial ha m o p i acy esul ing om AI sys ems collec ing la ge
olumes o pe sonal da a h ough moni o ing online beha iou and social media ac i i y
wi hou consen .
20
This da a can hen be used o p ofile and a ge indi iduals, modi ying
hei in o ma ion en i onmen wi hou hei awa eness and influencing how and whe he
hey engage in he democ a ic p ocess. Fu he mo e, combining gene a i e AI wi h AI-
powe ed social bo s could enable malicious ac o s o flood social media wi h an endless
s eam o messages ha dispa age and ein o ce s e eo ypes abou pa icula mino i y
g oups, co oding he mo al espec hey need o be ecognized as equal democ a ic in e -
locu o s.
21
Such in o ma ion ope a ions can spill o e in o he eal wo ld, con ibu ing o
hese g oups being u he ma ginalized and e en excluded om poli ics.
Simila hings can happen o g oups ha o ganize a ound poli ical in e es s a he
han s ic ly hei iden i ies, such as poli ical pa ies and social mo emen s. In addi ion
o la ge quan i ies o dispa aging in o ma ion, hei poli ical pa icipa ion can also be
jeopa dized by he comp omising ab ica ions c ea ed by deep akes. Poli ical ad e -
sa ies could use deep akes o depic each o he engaging in dis epu able beha iou ,
making con o e sial s a emen s, o pa icipa ing in c iminal ac i i ies while po ay-
ing hemsel es in a a ou able ligh . Finally, g oups and collec i es may find hei
oices d owned ou no because o he na u e o AI-gene a ed con en bu i s
massi e quan i y on social media and o he online pla o ms.
22
Sys emic ha m o democ acy
The sys emic ha m ha he widesp ead use o AI echnologies can inflic on democ-
acies e e s o he impai men o gene al socie al, in o ma ional, and poli ical con-
di ions in which democ a ic pa icipa ion and delibe a ion un old. Such ha m can
come in he o m o socie al and poli ical pola iza ion, agmen a ion, pe asi e dis-
us , and widesp ead poli ical apa hy and indiffe ence. These dis o ions need no
be p oduced by AI. They can a ise due o a a ie y o poli ical, cul u al, socie al,
and economic easons and, as we know, om digi al echnologies ha a e no augmen-
ed by AI. Howe e , AI-based echnologies, and especially gene a i e AI, h ea en o
exace ba e hem o his o ically unp eceden ed le els.
As al eady indica ed, igh s-based ha m and sys emic ha m o democ acy a e no
s ic ly sepa able. The eason o his lies in he ac ha he in ingemen o indi idual,
collec i e, and g oup igh s cons i u es one means by which AI can inflic sys emic ha m
on democ acy. To illus a e, conside he case o pe asi e AI p opaganda campaigns a -
ge ing a mino i y g oup on social media pla o ms. Such campaigns can simul aneously
e ode heg oup’s collec i e igh sand os e a clima e o in ensifiedsocie al adicaliza ion
and pola iza ion, hus de eloping om g oup igh s-ha m o sys emic ha m.
Howe e , he dis inc ion is none heless use ul, conside ing ha AI can unde mine
democ acy a a sys emic le el e en in he absence o explici iola ions o indi idual,
g oup, and collec i e igh s. S a egies o achie ing his can include sp eading a
ple ho a o conspi acy heo ies o ampli ying ex emis iews. Such campaigns can
lead o he o ma ion o isola ed, like-minded g oups ha a e inc easingly disconnec ed
om he b oade socie y, esul ing in socie al agmen a ion. A he same ime, he p o-
li e a ion o conspi acy heo ies can deepen he di ide be ween hose who subsc ibe o
hem and hose who ejec hem, he eby in ensi ying pola iza ion. This widening i
1072 J. CUPAĆAND M. SIENKNECHT
can make i mo e challenging o indi iduals and g oups o engage in cons uc i e dia-
logue and find common g ound, ul ima ely unde mining he democ a ic p ocess.
Mul iple scena ios can also be imagined in which he widesp ead use o AI can lead
o dis us and poli ical apa hy. Fo ins ance, hey may inad e en ly p io i ize con en
ha igge s s ong emo ional eac ions, such as ou age, ea , o disgus , leading o a
mo e nega i e and cynical pe cep ion o poli ics and poli icians. Using AI in mass su -
eillance may esul in pe cei ed o ac ual p i acy loss, making people mo e cau ious
abou exp essing hei opinions and engaging in poli ical ac i i ies. Las ly, AI-p o-
duced disin o ma ion and misin o ma ion can e ode us in adi ional knowledge
sou ces, such as uni e si ies, leading ci izens o ega d all u h claims wi h scep icism,
pe cei ing hem as poli ically mo i a ed.
Be o e we p oceed, i should be emphasized ha he li e a u e ecognizes he chal-
lenges o add essing and egula ing sys emic ha m. A p ima y conce n is he issue o
empo ali y. On he one hand, he e a e sys emic ha ms ha emain unan icipa ed and
hus elude egula ion. On he o he hand, he sys emic ha m we a e amilia wi h does
no a ise om a singula use o AI bu e ol es o e ime om mul iple deploymen s by
a ious ac o s and in combina ion wi h o he social ac o s. Consequen ly, his ype o
ha m is accumula i e, making i challenging o iden i y a di ec cause o de e mine he
esponsible pa y.
23
The EU’s esponse o AI ha m o democ acy: a c i ical su ey
Me hodology
How does he only sup ana ional o ganiza ion o i s kind comp ising 27 democ acies,
he EU, espond o he ou lined dange s posed by AI echnologies o democ acy? Wha
measu es is he EU aking o mee he challenge o egula ing a cons an ly e ol ing
echnology? In his sec ion, we su ey he EU’s policies and egula ions aimed a mi i-
ga ing igh s-based and sys emic ha m he widesp ead use o AI may cause o democ-
acy. Since he mid-1990s, he EU has egula ly in oduced policies, ac ion plans, codes
o conduc , and egula ions o suppo membe s a es’digi al ansi ion. To selec
documen s ele an o ou analysis, we fi s compiled a de ailed lis o he EU’s
digi al egula ions and policies by using he EU’s websi e and EUR-Lex sea ch unc ion
(see Appendix, Table 1). The esul ing lis is no exhaus i e. When i comes o non-
legal ins umen s, i only ea u es documen s wi h a di ec o indi ec connec ion o
AI, a he han co e ing digi aliza ion mo e b oadly.
To selec ele an documen s o ou analysis, in line wi h ou defini ion o democ-
acy, we iden ified documen s conce ned wi h igh s o poli ical pa icipa ion and
hose add essing he in eg i y o he in o ma ion en i onmen . F om he sho lis ed
documen s we hen selec ed o ou analysis hose add essing AI echnologies and
applica ions, specifically da a collec ion, p ocessing, p ofiling, a ge ing, social bo s,
deep- akes, and disin o ma ion campaigns, as well as o he echnologies ha could
b oadly be defined as AI.
This selec ion p ocess yielded ou legally binding documen s: The Gene al Da a
P o ec ion Regula ion (GDPR), he Digi al Se ices Ac (DSA), he A ificial In elli-
gence Ac (AIA), and he P oposal o a Regula ion on he T anspa ency and Ta ge ing
o Poli ical Ad e ising (TTPA). The GDPR is a p i acy law ha p o ec s EU esiden s’
da a by es ic ing i s unconsen ed collec ion and p ocessing, including o au oma ed
DEMOCRATIZATION 1073
decision-making. The DSA is a legisla i e amewo k o digi al se ices, aiming o
enhance online sa e y by comba ing disin o ma ion and ha m ul con en . The EU
AI Ac is he fi s legisla ion aimed a di ec ly egula ing he de elopmen and deploy-
men o AI sys ems wi hin he common ma ke . Finally, he TTPA is a p oposed law
designed o ha monize poli ical campaigning ac oss he con inen , including he use o
AI echnologies such as p ofiling and a ge ing.
The Eu opean Media F eedom Ac (EMFA) and he Digi al Ma ke s Ac (DMA) a e
examples o he documen s we excluded om ou analysis, al hough hey may appea
ele an a fi s glance. While he EMFA emphasizes democ a ic pa icipa ion, i does
no di ec ly add ess AI. Con e sely, he DMA men ions AI bu does no ocus on
democ acy as we define i he e in his a icle. As o non-binding legal and non-
legal documen s, we conside ed documen s in ou analysis ha ulfilled he same c i-
e ia. Thus, we included decla a ions, communica ions, whi e pape s, policies, and
codes o p ac ices ha deal wi h h ea s o democ acies and he influence o AI. Fo
a de ailed lis , see he bolded ex in he able loca ed in he Appendix.
To iden i y he specific EU egula o y and policy measu es in ended o p o ec he
con inen ’s democ acy om AI isks, we s a ed by induc i ely analysing he ou
legally binding documen s. By g ouping he a icles acco ding o hei simila i y, we
can show ha he measu es he EU is cu en ly aking o p o ec democ acy om
AI- ela ed ha m all in o ou b oad ca ego ies: (1) p ohibi ion, (2) anspa ency (sub-
ca ego ies: anspa ency o indi iduals, ope a ional anspa ency, epo ing anspa -
ency); (3) isk managemen ; and (4) educa ion (see Appendix, Table 2). In he
cou se o he analysis, wo mo e ca ego ies we e conside ed: socio- echnical design
o AI sys ems and he go e nance o he digi al sphe e. Howe e , we decided o
omi hem, conside ing hem as al eady con ained wi hin ou ou p ima y ca ego ies.
While specific a icles in he AIA, like A icle 15, explici ly equi e de elope s o
adhe e o ce ain design p inciples, we con end ha he bulk o he a icles wi hin
he p ohibi ion, anspa ency, and isk managemen ca ego ies also guide he design
o AI sys ems. These a icles, by dic a ing wha AI sys ems should o should no do
o a oid causing ha m, o asking de elope s o disclose ea u es o design, essen ially
se e as di ec i es o how hese sys ems should be made. Fo his eason, we ha e
op ed no o ea socio- echnical design as a sepa a e ca ego y, al hough we acknowl-
edge ha some a icles migh be mo e explici abou i . Simila ly, gi en ha go e n-
ance plays a c ucial ole in en o cing any EU egula ion, we see i as unnecessa y o
ea i as a dis inc ca ego y wi hin he con ex o digi al sphe e go e nance. In he
nex sec ion, we discuss each measu e, along wi h hei s eng hs and weaknesses in
educing igh s-based and sys emic AI ha ms o democ acy. In doing so, we le e age
insigh s om ecen esea ch in p ohibi ion, anspa ency, isk, and educa ion wi hin
he digi aliza ion field.
EU ins umen s agains AI ha m o democ acy
P ohibi ion
The EU’s mos a - eaching ins umen o mi iga e he ha m ul effec s o AI, including
ha m o democ acy, is he p ohibi ion o specific AI sys ems and p ac ices. Each o he
ou documen s examined – he GDPR, DSA, TTPA, and he AIA –includes such p o-
hibi ions. The GDPR bans p ocessing biome ic da a and da a ela ing o e hnic
1074 J. CUPAĆAND M. SIENKNECHT
While he EU’s ini ia i e o educa e i s ci izens on sa egua ding hei a ious igh s
in he ace o an inc easingly AI-d i en democ acy and a e ing sys emic ha m is com-
mendable, expec ing indi iduals o manage hei p i a e in o ma ion and disce n alse
in o ma ion consis en ly and collec i ely may be o e ly op imis ic. To begin wi h,
access o digi al educa ion is no equally a ailable. Mos EU campaigns a ge young
people in educa ional se ings, lea ing adul s and non-digi al na i es, who a e no
longe pa o he educa ional sys em and migh need AI cu icula ailo ed o hei cog-
ni i e le el, somewha o e looked.
59
Mo eo e , he EU has no sufficien ly add essed a
c ucial poin highligh ed in he li e a u e: aining AI de elope s on AI ha m o ensu e
he c ea ion o sys ems ha adhe e o he s anda ds o ai ness, accoun abili y, ans-
pa ency and e hics, and democ acy in gene al.
60
S ill, e en wi h ci izens and de elope s being highly educa ed abou AI isks and he
measu es a ailable o mi iga e hem, p o ec ing democ acy a bo h igh s-based and
sys emic le els is no gua an eed. While digi al li e acy is necessa y, i is no
sufficien . E en expe ienced social media mode a o s find iden i ying disin o ma ion
campaigns and in o ma ion p opaga ed by social bo s challenging. This challenge
g ows exponen ially wi h la ge language models capable o flooding he sys em wi h
misleading con en . Mo eo e , e en digi ally li e a e indi iduals migh become dis-
us ul and apa he ic i hey eel o e whelmed by he cons an need o diffe en ia e
be ween genuine in o ma ion and disin o ma ion and handling hei p i a e da a. I o-
nically, his could con ibu e o sys emic ha m o democ acy a he han p o ec ing
agains i .
Conclusion
In his s udy, we mapped he di e se measu es he EU has implemen ed o plans o
implemen o p o ec democ acy om he g owing ha m associa ed wi h AI echnol-
ogies. The mapping was guided by he dis inc ion we made be ween igh s-based and
sys emic AI ha m o democ acy, esul ing in ou ca ego ies o measu es: p ohibi ion,
anspa ency, isk managemen , and educa ion. As a mapping exe cise, ou analysis
ades an in-dep h examina ion o he EU egula ion o a bi d’s eye pe spec i e on
he egula o y landscape. Ye , his app oach enables us o in e se e al insigh s ha
could p omp a deepe analysis o hese measu es in he u u e.
We no e ha when we expand ou ocus beyond he AIA o o he binding and non-
binding EU documen s ha add ess o some deg ee he ha m o AI o democ acy, we
see ha he EU has de eloped a qui e obus oolbox o p o ec i e measu es, ca e ing o
bo h igh s-based and sys emic ha m. Especially in he ealm o sys emic ha m, which
exis ing li e a u e iden ifies as inadequa ely add essed, we map a a ie y o measu es
ha can be seen as effo s o ins i u e con inuous o e sigh o AI sys ems’de elopmen
and deploymen .
61
Howe e , despi e he inclusion o hese p o isions, AI s ill has significan po en ial
o p oduce sys emic ha m o democ acy. Based on his, we p opose ha he challenge
wi h AI sys emic ha m in he EU’s AI egula ion is no jus i s insufficien co e age, as
highligh ed by exis ing esea ch, bu also ha such ha m is inhe en ly difficul , pe haps
e en impossible, o effec i ely egula e.
Fi s , as unde sco ed h oughou he a icle, highly p o ec ed igh s and an AI-li -
e a e public can, pa adoxically, engende sys emic ha m o democ acy. Many people
consen ing o being p ofiled and a ge ed, e en wi h non-decep i e con en , can
DEMOCRATIZATION 1081
os e he c ea ion o echo chambe s, which, in u n, may lead o poli ical pola iza ion
and agmen a ion. Mo eo e , a public possessing a high le el o knowledge abou AI
p ac ices may g ow disillusioned wi h i s in eg a ion in o he democ a ic p ocess, con-
sequen ly becoming mo e dis us ul and poli ically apa he ic.
Second, he agg ega ed na u e o sys emic ha m, coupled wi h he ac ha i a ises
no solely om AI bu om i s in e wining wi h p e-exis ing social condi ions,
makes pinpoin ing i s exac causes challenging. Howe e , wi hou such p ecision,
i is difficul o es ablish pa ies accoun able o sys emic ha m o a h eshold a
which measu es a ge ing only i s AI componen would be effec i e. Fo example,
while social media companies con ibu e o inc easing poli ical pola iza ion, i is
unclea whe he elimina ing mis- and disin o ma ion om hese pla o ms would
alle ia e his issue o , impo an ly, wha amoun would need o be emo ed be o e
we see a posi i e shi .
Agains his backd op, we p opose wo a enues o u he esea ch: one analy i-
cal and he o he no ma i e. Analy ically, he e is a need o a ho ough examina ion
o he EU’s egula o y ins umen s o mi iga e he ha ms AI can cause o democ acy
a a undamen al le el. This examina ion should be mind ul o bo h igh s-based and
sys emic AI ha m as well as o he b oade egula o ylandscapeweou lined.Sc u-
iny o his ype could ac as a ca alys o egula o y imp o emen s and policy ec-
ommenda ions, which a e u gen ly needed gi en AI’s apid ad ancemen and
widesp ead adop ion. No ma i ely, i may be ime o acknowledge ha egula ion
and policy can go only so a and ha we need o s a hinking abou a no el pa a-
digm o democ acy in which AI would no be only a p oblem o be managed bu i s
in eg al componen .
62
No es
1. Smuha, “F om a ‘Race o AI’ o a ‘Race o AI Regula ion’.”
2. A e adop ing he A ificial In elligence Ac (AIA) on 13 h Ma ch 2024, he Eu opean Pa lia-
men issued co ec ions on 16 h Ap il 2024. This co ec ed e sion is ci ed in his pape and
ep esen s he final d a o he AI Ac , which is awai ing final app o al by he Council o he
Eu opean Union a he ime o w i ing his pape .
3. See: Hildeb and , “Algo i hmic Regula ion and he Rule o Law”; Kol , “Algo i hmic Black
Swans”; Smuha, “Beyond a Human Righ s-Based App oach o AI Go e nance”; Smuha,
“Beyond he Indi idual”; Uuk, “Manipula ion and he AI Ac ”; an de Sloo and an Schendel,
“P ocedu al Law o he Da a-D i en Socie y”; Yeung, “AI Go e nance by Human Righ s-Cen-
e ed Design.”
4. Eu opean Commission’s High-Le el Expe G oup on A ificial In elligence, “ADefini ion o
AI”,1.
5. C . Kaplan, “A ificial In elligence, Social Media, and Fake News,”153.
6. See: B kan, “A ificial In elligence and Democ acy”; Djeffal, “AI, Democ acy and he Law”;
Eubanks, Au oma ing Inequali y; Hinsch, “Diffe ences Tha Make a Diffe ence”; Ke yso a,
“A ificial In elligence and Disin o ma ion”; McSweeny, “Psychog aphics, P edic i e Analy ics,
A ificial In elligence & Bo s”;O’Neil, Weapons o Ma h Des uc ion; on Unge n-S e nbe g,
“Disc imina o y AI and he Law.”
7. See: B kan, “A ificial In elligence and Democ acy”; Djeffal, “AI, Democ acy, and he Law”;
Ienca, “On A ificial In elligence and Manipula ion”; Kaplan, “A ificial In elligence, Social
Media, and Fake News”; Ke yso a, “A ificial In elligence and Disin o ma ion”; König and
Wenzelbu ge , “Oppo uni y o Renewal o Dis up i e Fo ce?”; Milan and Agos i, “Pe sona-
lisa ion Algo i hms and Elec ions”; Mogaji e al.“Using AI o Pe sonalise Emo ionally Appeal-
ing Ad e isemen ”; Na ayanan, Unde s anding Social Media Recommenda ion Algo i hms;
Zacha y, “Digi al Manipula ion.”
1082 J. CUPAĆAND M. SIENKNECHT
8. See: B kan, “A ificial In elligence and Democ acy”;Diakopoulos,“Au oma ing he News”;Fe a a
e al., “TheRiseo SocialBo s”; Ga cía-O osa e al., “Algo i hms and Communica ion”; Kelle e al.
“Social Bo s in Elec ion Campaigns”;Shaoe al.,“The Sp ead o Low-C edibili y Con en .”
9. See: Habgood-Coo e, “Deep akes and he Epis emic Apocalypse”; Hamelee s e al., “You Won’
Belie e Wha They Jus Said!”; Jacobsen and Simpson, “The Tensions o Deep akes”; Wes e -
lund, “The Eme gence o Deep ake Technology”; Why e, “Deep ake News.”
10. Fo an o e iew see: Fa ahany, The Ba le o You B ain; Neuwi h, “P ohibi ed A ificial
In elligence P ac ices”; Neuwi h, The EU A ificial In elligence Ac .
11. See: Neuwi h, The EU A ificial In elligence Ac , 25–26 and 94.
12. McKay and Teno e, “Disin o ma ion as a Th ea o Delibe a i e Democ acy”; Teno e, “P o-
ec ing Democ acy om Disin o ma ion.”
13. Fo an o e iew, see: Wa en, “A P oblem-Based App oach o Democ a ic Theo y.”
14. Mansb idge e al., “A Sys emic App oach o Delibe a i e Democ acy”, 11-12.
15. Fo an o e iew, see: Shelby, “Socio echnical Ha ms o Algo i hmic Sys ems.”
16. Smuha, “Beyond a Human Righ s-Based App oach o AI Go e nance”; Smuha, “Beyond he
Indi idual”;Yeung, “AI Go e nance by Human Righ s–Cen e ed Design.”
17. Hildeb and , “Algo i hmic Regula ion and he Rule o Law”; Kol , “Algo i hmic Black Swans”;
Smuha, “Beyond he Indi idual”; Uuk, “Manipula ion and he AI Ac ”; an de Sloo and an
Schendel, “P ocedu al Law o he Da a-D i en Socie y.”
18. C . Junghe “A ificial In elligence and Democ acy”; König and Wenzelbu ge , “Oppo uni y
o Renewal o Dis up i e Fo ce.”
19. In “Beyond he Indi idual,”Smuha diffe en ia es be ween indi idual, collec i e, and socie al
ha m AI can cause. Con e sely, we adop a pe spec i e g ounded in he sys emic delibe a i e
defini ion o democ acy, which emphasizes pa icipa ion igh s o bo h indi iduals and
g oups, along wi h he sys emic condi ions necessa y o exe cising hese igh s. Consequen ly,
we offe a bina y ca ego iza ion o AI ha m: igh s-based and sys emic ha m. While he e a e
significan o e laps, ou classifica ion di e ges om Smuha’s by being ancho ed in a specific
unde s anding o democ acy a he han a b oad socie al iewpoin .
20. See o a discussion on he g oup igh o p i acy in he con ex o biomedical da a, Flo idi
2014.
21. C ., DiRes a e al., “The Tac ics & T opes.”
22. C ., Woolley and Guilbeaul , “Uni ed S a es.”
23. See: Smuha, “Beyond he Indi idual,”10; Ke nohan, And ew. “Accumula i e Ha ms.”; Kol ,
“Algo i hmic Black Swans.”
24. GDPR, Regula ion (EU) 2016/679, A icle 9.
25. DSA, Regula ion (EU) 2022/2065, A icle 26 and 28. DSA also p ohibi s designing online in e -
aces in a manipula i e way, A icle 25.
26. Eu opean Pa liamen and he Council o he Eu opean Union, Regula ion on he T anspa ency
and Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, A icle 18.
27. Some o hese p ohibi ions come wi h excep ions, o de ails see: Eu opean Pa liamen , Co i-
gendum, A ificial In elligence Ac , P9_TA(2024)0138, A icle 5. Fo de ailed analysis and c i-
icism o hese p ac ices see: Neuwi h, “P ohibi ed A ificial In elligence P ac ices,”; Neuwi h,
The EU A ificial In elligence Ac .
28. C . Smuha, “Beyond he Indi idual.”
29. Neuwi h, “P ohibi ed A ificial In elligence P ac ices,”4.
30. Jobin e al., “The Global Landscape o AI E hics Guidelines.”See also: Mike and C aw o d.
“Seeing Wi hou Knowing”; Go wa and Ash, “Democ a ic T anspa ency in he Pla o m
Socie y”; La sson, S e an, and F ed ik Hein z, “T anspa ency in A ificial In elligence.”
31. Albe Meije as quo ed in Diakopoulos, “T anspa ency,”198.
32. Heald, “Va ie ies o T anspa ency.”
33. GDPR, Regula ion (EU) 2016/679, A icle 12 and 13. T anspa ency measu es ela ing o he
handling o p i a e da a can also be ound in GDPR’s A icles 5, 6, 7, 8, 11, 14, 15, 16, 17,
18, 19, 20, 21, 22, and 34.
34. DSA, Regula ion (EU) 2022/2065, A icles 14, 17, 20, 26, 27, 28, 38, and 39; Eu opean Pa lia-
men and he Council o he Eu opean Union, Regula ion on he T anspa ency and Ta ge ing
o Poli ical Ad e ising, PE 90 2023 INIT, A icles 11, 12, 18, and 19; Eu opean Pa liamen ,
Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 13 and 50. I is impo an
DEMOCRATIZATION 1083
o highligh ha A icle 50 ex ends o AI unde a ee and open-sou ce licence. Some c i ics
iew his as an insufficien measu e o his ype o AI, which is o en ega ded as pa icula ly
isky o i s po en ial o ha m democ acy, including ia he widesp ead dissemina ion o dis-
in o ma ion. See, e.g.: Giannaccini and Kleineidam, How he EU’s So Touch on Open-Sou ce
AI Opens he Doo o Disin o ma ion.
35. See: And ada e al., “Va ie ies o T anspa ency”; Diakopoulos, “T anspa ency”; Walmsley,
“A ificial In elligence and he Value o T anspa ency.”
36. GDPR, Regula ion (EU) 2016/679, A icle 25, 40, 41, and 42.; DSA, Regula ion (EU) 2022/2065,
A icles 44 o 77; Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA
(2024)0138, A icles 10, 11, 12, 13, 15, 17, 18, 44, 47, 48, 49 and 77.
37. Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 13
and 53.
38. GDPR, Regula ion (EU) 2016/679, A icle 30 and 32; DSA, Regula ion (EU) 2022/2065,
A icles 37 and 40; Eu opean Pa liamen and he Council o he Eu opean Union, Regula ion
on he T anspa ency and Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, A icles 9 and 19;
Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 12,
19, 53, and 55.
39. DSA, Regula ion (EU) 2022/2065, A icles 10, 15, 24, and 42.
40. Eu opean Pa liamen and he Council o he Eu opean Union, Regula ion on he T anspa ency
and Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, A icles 14, 16, 17, and 20.
41. Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 20,
27, 52, 77, and 91.
42. Ananny and C aw o d, “Seeing Wi hou Knowing.”
43. GDPR, Regula ion (EU) 2016/679, A icles 35 and 36; DSA, Regula ion (EU) 2022/2065,
A icle 34; Eu opean Pa liamen and he Council o he Eu opean Union, Regula ion on he
T anspa ency and Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, A icle 19.
44. Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 27
and 43 and Annex III(8).
45. Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 57
and 60.
46. GDPR, Regula ion (EU) 2016/679, 33; DSA, Regula ion (EU) 2022/2065, A icles 16, 18, 22,
and 23; Eu opean Pa liamen and he Council o he Eu opean Union, Regula ion on he
T anspa ency and Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, A icle 15; Eu opean
Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icle 9, 14, 55, 72,
73, 74, 75, 76, 89, and 90.
47. Remedial and mi iga ion a icles can be ei he independen o a pa o some o he a icle. Fo
example, see: DSA, Regula ion (EU) 2022/2065, A icles 9, 16, 18, 22, 23, 35, 36, and 48; Eu o-
pean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icles 15, 20,
55, and 57.
48. The GDPR es ablishes he Da a P o ec ion Office , Lead Supe iso y Au ho i y, Eu opean Da a
P o ec ion Boa d, and Na ional Da a P o ec ion Au ho i ies; he DSA es ablishes he Digi al
Se ices Coo dina o , Eu opean Boa d o Digi al Se ices; and he AIA es ablishes he AI
Office, Eu opean A ificial In elligence Boa d, Ad iso y Fo um, Scien ific Panel o Indepen-
den Expe s, No i ying Au ho i y, Ma ke Su eillance Au ho i y, and Con o mi y Assessmen
Body.
49. Eu opean Commission, Commission is Ga he ing Views on D a DSA Guidelines o Elec ion
In eg i y.
50. Aus in, “The DSA now Applies in Full.”
51. Veale and Bo gesius, “Demys i ying he D a Eu A ificial In elligence Ac ,”105.
52. Edwa ds, Regula ing AI in Eu ope.
53. Eu opean Commission, COM (2020) 790 final, 3.
54. See: h ps://ec.eu opa.eu/eu os a /cache/me ada a/en/isoc_sk_dskl_i21_esmsip2.h m
55. Eu opean Commission, S anda d Eu oba ome e 98, QF8.1, T150.
56. On us wo hiness o sou ces, see: Eu opean Commission, COM (2020) 790, 23. On sha ing
he in o ma ion, see: Eu opean Decla a ion on Digi al Righ s and P inciples o he Digi al
Decade (2023/C 23/01); The S eng hened Code o P ac ice on Disin o ma ion (2022 COM
(2020) 624 final); Tackling Online Disin o ma ion: A Eu opean App oach (COM (2018) 236
1084 J. CUPAĆAND M. SIENKNECHT
final); Whi e Pape on A ificial In elligence: A Eu opean app oach o excellence and us
(COM (2020) 65 final).
57. Eu opean Pa liamen , Co igendum, A ificial In elligence Ac , P9_TA(2024)0138, A icle 3
(56).
58. Eu opean Commission, COM (2020) 790 final, 3.
59. Chan, “A Comp ehensi e AI Policy Educa ion F amewo k”; Kaplan, “A ificial In elligence,
Social Media, and Fake News”; Laupichle e al., “A ificial In elligence Li e acy in Highe
and Adul Educa ion”; Lee, “Fake News, Phishing, and F aud”; Yang, “A ificial In elligence
Educa ion o Young Child en”.
60. Bogina e al. “Educa ing So wa e and AI S akeholde s”; Bo ens ein and Howa d, “Eme ging
Challenges in AI”; Schiff,“Educa ion o AI, no AI o Educa ion.”
61. C . Smuha, “Beyond he Indi idual,”16-23.
62. See: Susskind, On F eedom and Democ acy; Coeckelbe ghm, Why AI Unde mines Democ acy.
Acknowledgmen
The au ho s wish o exp ess hei g a i ude o Yushu Soon o he assis ance and pa ience du ing he
li e a u e e iew o his a icle, and o he e iewe s o hei insigh ul and cons uc i e commen s.
Disclosu e s a emen
No po en ial conflic o in e es was epo ed by he au ho (s).
No es on con ibu o s
Jelena Cupaćis a Pos -Doc o al Resea ch Fellow a he WZB Be lin Social Science Cen e . She holds a
PhD om he Eu opean Uni e si y Ins i u e (EUI) in Flo ence. He esea ch explo es he ans o m-
a ion o in e na ional o ganiza ions, he global an i-gende mo emen , and effo s o go e n AI’s
impac on democ acy.
Mi ja Sienknech is a pos doc o al esea che a he Eu open New School o Digi al S udies/ Eu opean
Uni e si y Viad ina. He cu en esea ch ocuses on he ans o ma ion o wa h ough AI echnol-
ogies, he gene al impac o AI sys ems on democ a ic socie ies, and ela ed e hical ques ions.
ORCID
Jelena Cupaćh p://o cid.o g/0000-0002-7471-7624
Mi ja Sienknehc h p://o cid.o g/0000-0001-5217-1432
Bibliog aphy
Ananny, Mike, and Ka e C aw o d. “Seeing Wi hou Knowing: Limi a ions o he T anspa ency Ideal
and i s Applica ion o Algo i hmic Accoun abili y.”New Media & Socie y 20, no. 3 (2018): 973–989.
And ada, Glo ia, Robe W. Clowes, and Paul R. Sma . “Va ie ies o T anspa ency: Explo ing Agency
wi hin AI Sys ems.”AI & Socie y 38, no. 4 (2023): 1321–1331.
Bogina, Ve onika, Alan Ha man, Ts i Kuflik, and A i al Shulne -Tal. “Educa ing So wa e and AI
S akeholde s abou Algo i hmic Fai ness, Accoun abili y, T anspa ency and E hics.”
In e na ional Jou nal o A ificial In elligence in Educa ion 32, no. 3 (2021): 808–833.
Bo ens ein, Jason, and Ayanna Howa d. “Eme ging Challenges in AI and he Need o AI E hics
Educa ion.”AI and E hics 1, no. 1 (2021): 61–65.
B kan, Maja. “A ificial In elligence and Democ acy: The Impac o Disin o ma ion,: Social Bo s and
Poli ical Ta ge ing.”Delphi 1, no. 2 (2019): 66–71.
Chadha, Anupama, Vaibha Kuma , Sonu Kashyap, and Mayank Gup a. “Deep ake: An O e iew.”In
P oceedings o Second In e na ional Con e ence on Compu ing, Communica ions, and Cybe -
DEMOCRATIZATION 1085
Secu i y, edi ed by P adeep Kuma Singh, Sławomi T. Wie zchoń, Sudeep Tanwa , Ma ia Ganzha,
and Joel J. P. C. Rod igues, 557–566. Singapo e: Sp inge Singapo e, 2021.
Chan, Cecilia Ka Yuk. “A Comp ehensi e AI Policy Educa ion F amewo k o Uni e si y Teaching
and Lea ning.”In e na ional Jou nal o Educa ional Technology in Highe Educa ion 20, no. 1
(2023): 1–25.
Coeckelbe gh, Ma k. Why AI Unde mines Democ acy and Wha o Do abou I . Camb idge: Poli y
P ess, 2024.
Da is, Aus in. “The DSA Now Applies in Full: Wha Can we Expec ?”, Democ acy Repo ing
In e na ional, 21 Feb ua y 2024.: h ps://democ acy- epo ing.o g/en/office/EU/publica ions/dsa-
day- aking-s ock-and-looking-ahead- o-a-supe -elec ion-yea .
Diakopoulos, Nicholas. “T anspa ency.”In The Ox o d Handbook o E hics o AI, edi ed by Ma kus D.
Dubbe , F ank Pasquale, and Suni Das, 197–213. New Yo k: Ox o d Uni e si y P ess, 2020.
Diakopoulos, Nicholas. Au oma ing he news: How algo i hms a e ew i ing he media. Camb idge,
Massachuse s: Ha a d Uni e si y P ess, 2019.
DiRes a, Renee, K is Shaffe , Becky Ruppel, Da id Sulli an, Robe Ma ney, Ryan Fox, Jona han
Alb igh , and Ben Johnson. “The Tac ics & T opes o he In e ne Resea ch Agency.”Cong ess
o he Uni ed S a es: U.S. Sena e Documen , 2019. h ps://digi alcommons.unl.edu/cgi/
iewcon en .cgi?a icle=1003&con ex =sena edocs.
Djeffal, Ch is ian. “AI, Democ acy, and he Law.”In The Democ a iza ion o A ificial In elligence,
edi ed by And eas Sudmann, 255–284. Biele eld: T ansc ip , 2019.
DSA, Digi al Se ices Ac ,: Regula ion (EU) 2022/2065, Oc obe 19, 2022.
Edwa ds, Lilian. Regula ing AI in Eu ope: Fou P oblems and Fou Solu ions. London: Ada Lo elace
Ins i u e, 2022.
Eubanks, Vi ginia. Au oma ing Inequali y: How High-Tech Tools P ofile, Police, and Punish he Poo .
New Yo k: S . Ma in’s P ess, 2018.
Eu opean Commission’s High-Le el Expe G oup on A ificial In elligence: A Defini ion o AI: Main
Capabili ies and Scien ific Disciplines, B ussels, 18 Decembe 2018.
Eu opean Commission. “Communica ion om he Commission o he Eu opean Pa liamen , he
Council, he Eu opean Economic and Social Commi ee and he Commi ee o he Regions on
he Eu opean Democ acy Ac ion Plan,”COM (2020) 790 final, Decembe 12, 2020.
Eu opean Commission. S anda d Eu oba ome e 98 - Win e 2022-2023. Public opinion in he
Eu opean Union, 2023. h ps://eu opa.eu/eu oba ome e /su eys/de ail/2872.
Eu opean Commission. “Commission is Ga he ing Views on D a DSA Guidelines o Elec ion
In eg i y,”P ess Release, 8 Feb ua y 2024. h ps://digi al-s a egy.ec.eu opa.eu/en/news/commiss
ion-ga he ing- iews-d a -dsa-guidelines-elec ion-in eg i y.
Eu opean Pa liamen . “Co igendum o he Legisla i e esolu ion o 13 Ma ch 2024 on he posi ion o
he Eu opean Pa liamen adop ed a fi s eading wi h a iew o he adop ion o Regula ion (EU)
2024/ ……o he Eu opean Pa liamen and o he Council laying down ha monised ules on
a ificial in elligence and amending Regula ions (EC) No 300/2008, (EU) No 167/2013, (EU) No
168/2013, (EU) 2018/858, (EU) 2018/1139, and (EU) 2019/2144 and Di ec i es 2014/90/EU,
(EU) 2016/797, and (EU) 2020/1828 (A ificial In elligence Ac ), P9_TA(2024)0138, (COM
(2021)0206 –C9-0146/2021–2021/0106(COD))”.
Eu opean Pa liamen and he Council o he Eu opean Union: Regula ion on he T anspa ency and
Ta ge ing o Poli ical Ad e ising, PE 90 2023 INIT, Feb ua y 29, 2024.
Fa ahany, Ni a A. The Ba le o You B ain: De ending he Righ o Think F eely in he Age o
Neu o echnology. New Yo k: S . Ma in’s P ess, 2023.
Fe a a, Emilio, Onu Va ol, Clay on Da is, Filippo Mencze , and Alessand o Flammini. “The Rise o
Social Bo s.”Communica ions o he ACM 59, no. 7 (2016): 96–104.
Flo idi, Luciano. “Open Da a, Da a P o ec ion, and G oup P i acy.”Philosophy & Technology 27
(2014): 1–3.
Ga cía-O osa, Be a, João Cana ilhas, and Jo ge Vázquez-He e o. “Algo i hms and Communica ion:
A Sys ema ized Li e a u e Re iew.”Comunica 31, no. 74 (2023): 9–21.
GDPR. Gene al Da a P o ec ion Regula ion, Regula ion (EU) 2016/679, Ap il 4, 2016.
Giannaccini, F ancesca, and Tobias Kleineidam. “How he EU’s So Touch on Open-Sou ce AI
Opens he Doo o Disin o ma ion,”The Pa liamen , 16 Ap il 2024. h ps://www. hepa liamen
magazine.eu/news/a icle/how- he-eus-so - ouch-on-opensou ce-ai-opens- he-doo - o-disin o
ma ion.
1086 J. CUPAĆAND M. SIENKNECHT
Go wa, Robe , and Timo hy Ga on Ash. “Democ a ic T anspa ency in he Pla o m Socie y.”
In Social Media and Democ acy: The S a e o he Field and P ospec s o Re o m,
edi ed by Na haniel Pe sily, and Joshua A. Tucke , 286–312. Camb idge: Camb idge Uni e si y
P ess, 2020.
Habgood-Coo e, Joshua. “Deep akes and he Epis emic Apocalypse.”Syn hese 201, no. 103 (2023):
1–23.
Hamelee s, Michael, Toni G.L.A. an de Mee , and Tom Dobbe . “You Won’ Belie e Wha They Jus
Said! The Effec s o Poli ical Deep akes Embedded as Vox Populi on Social Media.”Social Media +
Socie y 8, no. 3 (2022): 1–12.
Hamon, Ronan, Hen ik Junklewi z, and Ignacio Sanchez. “Robus ness and Explainabili y o A ificial
In elligence.”Publica ions Office o he Eu opean Union 207 (2020.
Heald, Da id Albe . “Va ie ies o T anspa ency.”In T anspa ency: The Key o Be e Go e nance?
P oceedings o he B i ish Academy, edi ed by Ch is ophe Hood, and Da id Heald, 25–43.
Ox o d: Ox o d Uni e si y P ess, 2006.
Hildeb and , Mi eille. “Algo i hmic Regula ion and he Rule o Law.”Philosophical T ansac ions o he
Royal Socie y A: Ma hema ical, Physical and Enginee ing Sciences 376, no. 2128 (2018): 20170355.
h ps://doi.o g/10.1098/ s a.2017.0355.
Hinsch, Wil ed. “Diffe ences Tha Make a Diffe ence: Compu a ional P ofiling and Fai ness o
Indi iduals.”In The Camb idge Handbook o Responsible A ificial In elligence, edi ed by Silja
Voeneky, Philipp Kellmeye , Oli e Muelle , and Wol am Bu ga d, 229–251. Camb idge:
Camb idge Uni e si y P ess, 2022.
Ienca, Ma cello. “On A ificial In elligence and Manipula ion.”Topoi 42 (2023): 1–10.
Jacobsen, Benjamin N., and Jill Simpson. “The Tensions o Deep akes.”In o ma ion, Communica ion
& Socie y (2023): 1–15. h ps://doi.o g/10.1080/1369118X.2023.2234980.
Jobin, Anna, Ma cello Ienca, and Effy Vayena. “The Global Landscape o AI E hics Guidelines.”
Na u e Machine In elligence 1, no. 9 (2019): 389–399.
Junghe , And eas. “A ificial In elligence and Democ acy: A Concep ual F amewo k.”Social Media +
Socie y 9, no. 3 (2023): 1–14.
Kaplan, And eas. “A ificial In elligence, Social Media, and Fake News: Is This he End o
Democ acy?”In Digi al T ans o ma ion in Media & Socie y, edi ed by Ayşen Akko Gül, Yıldız
Dilek E ü k, and Paul Elme , 149–161. Is anbul: Is anbul Uni e si y P ess, 2020.
Kelle , Tobias R., and Ul ike Klinge . “Social Bo s in Elec ion Campaigns: Theo e ical,: Empi ical, and
Me hodological Implica ions.”Poli ical Communica ion 36, no. 1 (2019): 171–189.
Ke nohan, And ew. “Accumula i e Ha ms and he In e p e a ion o he Ha m P inciple.”Social
Theo y and P ac ice 19, no. 1 (1993): 51–72.
Ke yso a, Ka a ina. “A ificial In elligence and Disin o ma ion: How AI Changes he Way
Disin o ma ion Is P oduced: Dissemina ed, and Can Be Coun e ed.”Secu i y and Human Righ s
29, no. 1–4 (2018): 55–81.
Kol , Noam. “Algo i hmic Black Swans.”Washing on Uni e si y Law Re iew 101, ( o hcoming).
König, Pascal D., and Geo g Wenzelbu ge . “Oppo uni y o Renewal o Dis up i e Fo ce? How
A ificial In elligence Al e s Democ a ic Poli ics.”Go e nmen In o ma ion Qua e ly 37, no. 3
(2020): 101489.
La sson, S e an, and F ed ik Hein z. “T anspa ency in A ificial In elligence.”In e ne Policy Re iew 9,
no. 2 (2020): 1–16.
Laupichle , Ma hias Ca l, Alexand a As e , Jana Schi ch, and Tobias Raupach. “A ificial In elligence
Li e acy in Highe and Adul Educa ion: A Scoping Li e a u e Re iew.”Compu e s and Educa ion:
A ificial In elligence 3, no. 100101 (2022): 1–15.
Lee, Nicole M. “Fake News,: Phishing, and F aud: A Call o Resea ch on Digi al Media Li e acy
Educa ion Beyond he Class oom.”Communica ion Educa ion 67, no. 4 (2018): 460–466.
Mansb idge, Jane, James Bohman, Simone Chambe s, Thomas Ch is iano, A chon Fung, John
Pa kinson, Dennis F. Thompson, and Ma k E. Wa en. “A Sys emic App oach o Delibe a i e
Democ acy.”In Delibe a i e Sys ems: Delibe a i e Democ acy a he La ge Scale, edi ed by Jane
Mansb idge, and John Pa kinson, 1–26. Camb idge: Camb idge Uni e si y P ess, 2012.
McKay, Spence , and Ch is Teno e. “Disin o ma ion as a Th ea o Delibe a i e Democ acy.”Poli ical
Resea ch Qua e ly 74, no. 3 (2021): 703–717.
McSweeny, Te ell. “Psychog aphics,: P edic i e Analy ics, A ificial In elligence, & Bo s: Is he FTC
Keeping Pace?”Geo ge own Law Technology Re iew 2 (2018): 514.
DEMOCRATIZATION 1087
Milan, S e ania, and Claudio Agos i. “Pe sonalisa ion Algo i hms and Elec ions: B eaking F ee o he
Fil e Bubble.”In e ne Policy Re iew, 7 Feb ua y (2019). h ps://policy e iew.in o/a icles/news/
pe sonalisa ion-algo i hms-and-elec ions-b eaking- ee-fil e -bubble/1385.
Mogaji,Emmanuel,SundayOlaleye,andDandisonUkpabi.“Using AI o Pe sonalise Emo ionally Appealing
Ad e isemen .”In Digi al and Social Media Ma ke ing: Eme ging Applica ions and Theo e ical
De elopmen s, edi ed by N ipend a P. Rana, Emma L. Slade, Ganesh P. Sahu, Ha ice Kizgin, Ni ish
Singh,Bidi Dey,AnabelGu ie ez,andYogeshK.Dwi edi,137–150. Cham: Sp inge , 2020.
Na ayanan, A ind. Unde s anding Social Media Recommenda ion Algo i hms. New Yo k: Kings Fi s
Amendmen Ins i u e a Columbia Uni e si y, 2023.
Neuwi h, Ros am J. “P ohibi ed A ificial In elligence P ac ices in he P oposed EU A ificial
In elligence Ac (AIA).”Compu e Law & Secu i y Re iew 48 (2023): 105798.
Neuwi h, Ros am J. The EU A ificial In elligence Ac : Regula ing Subliminal AI Sys ems. New Yo k:
Rou ledge, 2023.
O’Neil, Ca hy. Weapons o Ma h Des uc ion: How Big Da a Inc eases Inequali y and Th ea ens
Democ acy. New Yo k: C own, 2017.
Susskind, Jamie. The Digi al Republic: On F eedom and Democ acy in he 21s Cen u y. London:
Bloomsbu y Publishing, 2022.
Schiff,Daniel.“Educa ion o AI,: no AI o Educa ion: The Role o Educa ion and E hics in Na ional AI
Policy S a egies.”In e na ional Jou nal o A ificial In elligence in Educa ion 32, no. 3 (2022): 527–563.
Shao, Chengcheng, Gio anni Luca Ciampaglia, Onu Va ol, Kai-Cheng Yang, Alessand o Flammini,
and Filippo Mencze . “The Sp ead o Low-C edibili y Con en by Social Bo s.”Na u e
Communica ions 9, no. 1 (2018): 1–9.
Shelby, Renee, Shalaleh Rismani, Ka h yn Henne, Ajung Moon, Nega Ros amzadeh, Paul Nicholas,
N’Mah Yilla-Akba i, e al. “Socio echnical Ha ms o Algo i hmic Sys ems: Scoping a Taxonomy o
Ha m Reduc ion.”In P oceedings o he 2023 AAAI/ACM Con e ence on AI, E hics, and Socie y
(2023): 723–741.
Smuha, Na halie A. “Beyond a Human Righ s-Based App oach o AI Go e nance: P omise,: Pi alls,
Plea.”Philosophy & Technology 34, no. Suppl 1 (2021): 91–104. h ps://doi.o g/10.1007/s13347-
020-00403-w.
Smuha, Na halie A. “Beyond he Indi idual: Go e ning AI’s Socie al Ha m.”In e ne Policy Re iew 10,
no. 3 (2021): 1–32.
Smuha, Na halie A. “F om a ‘Race o AI’ o a ‘Race o AI Regula ion’: Regula o y Compe i ion o
A ificial In elligence.”Law, Inno a ion and Technology 13, no. 1 (2021): 57–84.
Teno e, Ch is. “P o ec ing Democ acy om Disin o ma ion: No ma i e Th ea s and Policy
Responses.”The In e na ional Jou nal o P ess/Poli ics 25, no. 3 (2020): 517–537.
Uuk, Ris o. “Manipula ion and he AI Ac .”The Fu u e o Li e Ins i u e (2022).
an de Sloo , Ba , and Sascha an Schendel. “P ocedu al Law o he Da a-D i en Socie y.”
In o ma ion & Communica ions Technology Law 30, no. 3 (2021): 304–332.
Veale, M., and F. Z. Bo gesius. “Demys i ying he D a EU A ificial In elligence Ac —Analysing he
Good, he Bad, and he Unclea Elemen s o he P oposed App oach.”Compu e Law Re iew
In e na ional 22, no. 4 (2021): 97–112.
on Unge n-S e nbe g, An je. “Disc imina o y AI and he Law: Legal S anda ds o Algo i hmic
P ofiling.”In The Camb idge Handbook o Responsible A ificial In elligence, edi ed by Silja
Voeneky, Philipp Kellmeye , Oli e Muelle , and Wol am Bu ga d, 252–278. Camb idge:
Camb idge Uni e si y P ess, 2022.
Walmsley, Joel. “A ificial In elligence and he Value o T anspa ency.”AI & Socie y 36, no. 2 (2021):
585–595.
Wa en, Ma k E. “A P oblem-Based App oach o Democ a ic Theo y.”Ame ican Poli ical Science
Re iew 111, no. 1 (2017): 39–53.
Wes e lund, Mika. “The Eme gence o Deep ake Technology: A Re iew.”Technology Inno a ion
Managemen Re iew 9, no. 11 (2019): 39–52.
Why e, Ch is ophe . “Deep ake News: AI-enabled Disin o ma ion as a Mul i-Le el Public Policy
Challenge.”Jou nal o Cybe Policy 5, no. 2 (2020): 199–217.
Woolley, Samuel C., and Douglas Guilbeaul . “Uni ed S a es: Manu ac u ing Consensus Online.”In
Compu a ional P opaganda: Poli ical Pa ies, Poli icians, and Poli ical Manipula ion on Social
Media, edi ed by Samuel C. Woolley, and Philip N. Howa d, 185–211. Ox o d: Ox o d
Uni e si y P ess, 2018.
1088 J. CUPAĆAND M. SIENKNECHT
Yang, Weipeng. “A ificial In elligence Educa ion o Young Child en: Why,: Wha , and How in
Cu iculum Design and Implemen a ion.”Compu e s and Educa ion: A ificial In elligence 3, no.
100061 (2022): 1–7.
Yeung, Ka en, And ew Howes, and Ganna Pog ebna. “AI Go e nance by Human Righ s–Cen e ed
Design, Delibe a ion, and O e sigh .”In The Ox o d Handbook o E hics o AI,edi edbyMa kus
D. Dubbe , F ank Pasquale, and Suni Das, 77–106. New Yo k: Ox o d Uni e si y P ess, 2020.
Zacha y, G. Pascal. “Digi al Manipula ion and he Fu u e o Elec o al Democ acy in he US.”IEEE
T ansac ions on Technology and Socie y 1, no. 2 (2020): 104–112.
Appendix
Table 1. Binding and Non-Binding Documen s o he Eu opean Union Rela ing o Digi aliza ion.
Binding legal ins umen s
Regula ion and
P oposals o
Regula ion
Gene al Da a P o ec ion Regula ion (GDPR)
Digi al Se ices Ac (DSA)
A ificial In elligence Ac (AIA)
T anspa ency and Ta ge ing o Poli ical Ad e ising (TTPA)
Digi al Ma ke s Ac (DMA)
Media F eedom Ac
Da a Go e nance Ac
Cybe secu i y Ac
Regula ion on he P o ec ion o Na u al Pe sons wi h Rega d o he P ocessing o Pe sonal
Da a by he Union Ins i u ions, Bodies, Offices, and Agencies and on he F ee Mo emen o
Such Da a
Gigabi In as uc u e Ac
Eu opean Digi al Iden i y
The Eu opean Heal h Da a Space
Da a Ac
Chips Ac
P oposal o Regula ion on he Digi aliza ion o Judicial Coope a ion and Access o Jus ice
in C oss-Bo de Ci il, Comme cial and C iminal Ma e s, and Amending Ce ain Ac s in he
Field o Judicial Coope a ion
P oposal o Amending Regula ion (EU) No 904/2010 as Rega ds he VAT Adminis a i e
Coope a ion A angemen s Needed o he Digi al Age
Di ec i es Da a P o ec ion Law En o cemen Di ec i e
e-P i acy Di ec i e
Di ec i e on comba ing he sexual abuse and sexual exploi a ion o child en and child
po nog aphy
Audio isual Media Se ices Di ec i e (AVMSD)
Di ec i e conce ning measu es o a high common le el o secu i y o ne wo k and
in o ma ion sys ems ac oss he Union
Non-Binding Legal ins umen s
Decla a ions Decla a ion on Eu opean Digi al Righ s and P inciples o he Digi al Decade
Decla a ion: Coope a ion on A ificial In elligence
Communica ions Communica ion on p o ec ing elec ion in eg i y and p omo ing democ a ic
pa icipa ion
Tackling Online Disin o ma ion: a Eu opean App oach
Ac ion Plan agains Disin o ma ion
A ificial In elligence o Eu ope
Coo dina ed Plan on A ificial In elligence
Building T us in Human Cen ic A ificial In elligence
Fos e ing a Eu opean App oach o A ificial In elligence (+Annex)
P o ec ing Elec ion In eg i y and P omo ing Democ a ic Pa icipa ion
Digi aliza ion o Jus ice in he Eu opean Union
EU Policy on Cybe De ence
Media and Audio isual Ac ion Plan
Eu opean S a egy o Da a
Digi al Educa ion Ac ion Plan 2021–2027
A Chips Ac o Eu ope
DEMOCRATIZATION 1089
Digi aliza ion o Jus ice in he Eu opean Union
Union o Equali y: S a egy o he Righ s o Pe sons wi h Disabili ies (2021-2030)
A Digi al Decade o Child en and You h: The New Eu opean S a egy o a Be e In e ne
o Kids (BIK+)
A Eu opean Heal h Da a Space: Ha nessing he Powe o Heal h Da a o People, Pa ien s
and Inno a ion
A Eu opean s a egy on Coope a i e In elligen T anspo Sys ems, a Miles one Towa ds
Coope a i e, Connec ed and Au oma ed Mobili y
FinTech Ac ion plan: Fo a mo e compe i i e and inno a i e Eu opean financial sec o
Whi e pape s Whi e Pape on A ificial In elligence: A Eu opean App oach o Excellence and T us
Non-legal ins umen s
Policies,
P og ammes,
Ini ia i es
Shaping Eu ope’s Digi al Fu u e
Eu opean Democ acy Ac ion Plan
Digi al Educa ion Ac ion Plan 2021–2027
Eu opean G een Digi al Coali ion
Nex Gene a ion In e ne Ini ia i e (NGI)
Policy Guidance on AI o Child en
Global Ga eway
Ge Digi al Ini ia i e
Codes o
P ac ice
The S eng hened Code o P ac ice on Disin o ma ion
Table 2. Classifica ion o he EU’s measu es o mi iga e AI ha m o democ acy*.
T anspa ency
EU
documen s P ohibi ion
T anspa ency
o indi idual
Ope a ional
anspa ency
Repo ing
T anspa ency
Risk
managemen Educa ion**
GDPR A icles 9 A icles 5, 6, 7,
8, 11, 12, 13,
14, 15, 16, 17,
18, 19, 20, 21,
22, 34
A icles 25, 30,
32, 40, 41, 42
A icle 31, 47 A icles 33, 35,
36
A icle 57
DSA A icles 25,
26, 28
A icles 14, 17,
20, 26, 27, 28,
38, 39
A icles 37, 40,
44, 45, 46, 47,
48, 49, 50, 51,
52, 53, 54, 55,
56, 57, 58, 59,
60, 61, 62, 63,
64, 65, 66, 67,
68, 69, 70, 71,
72, 73, 74, 75,
76, 77
A icles 10,
15, 24, 42
A icles 9, 16,
18, 22, 23,
34, 35, 36,
48
AIA A icle 5 A icles 13, 50 A icles 10, 11,
12, 13, 15, 17,
18, 19, 44, 47,
48, 49, 53, 55,
77
A icle 20, 27,
52, 77, 91
A icles 9, 14,
15, 20, 27,
43, 55, 57,
60, 72,
73,74, 75,
76, 89, 90
A icle 3
TTPA A icle 18 A icles 11, 12,
18, 19
A icle 9, 19 A icles 14,
16, 17, 20
A icle 15, 19
*The able selec i ely highligh s a icles mos pe inen o igh s-based and sys emic AI ha m o democ acy,
excluding hose wi h lesse ele ance. A icles may c oss se e al ca ego ies, eflec ing hei in e connec ed
oles in he documen s.
** The EU deals wi h educa ion- ela ed measu es p ima ily in i s non-binding legal ins umen s and non-legal
ins umen s while egula ions only spo adically co e his aspec .
1090 J. CUPAĆAND M. SIENKNECHT
