Abstract
When the European Union (EU) adopted its AI Act (AIA) in 2024, hopes were high that EU rules would diffuse globally through a “Brussels Effect.” We investigate how structural and contextual factors in the field of AI governance affect the likelihood and transformative potential of a Brussels Effect there. Because the AIA is still too young to allow a straightforward ex post analysis, we combine two inferential strategies: we compare AI as a governance challenge to other cases of successful EU rule export, identifying conditions that promote or obstruct global EU rule export. And we empirically canvas the rule-setting dynamics that have emerged since the AIA’s first legislative draft in 2021. Against the initially optimistic tenor of Brussels policy discourse from the time of the AIA’s adoption, we advance four reasons to be skeptical of big EU influence: first, the EU’s regulatory capacity is constrained by informational disadvantages, reliance on external standard-setters, and enforcement delays. Second, competitive pressures have led to pre-emptive concessions, watering down EU rules before they could travel. Third, the relative vagueness of EU rules means that de factor harmonization of rules and practices across borders remains shallow at best. Fourth, rule diffusion in AI is above all meaningful when it reaches countries that themselves are leading AI powers, notably the United States and China, and it is precisely there that EU influence seems negligible so far.
Similar content being viewed by others
1 Introduction
AI presents sundry societal and political challenges – from biased decisionmaking (Benjamin, 2019) and deleterious effects on democracy (Simons, 2023) to labor market disruption (Acemoğlu & Johnson, 2023) and environmental degradation (de Vries, 2023) – that make effective regulation essential. When the European Union (EU) adopted the AI Act (AIA) in March 2024, it hoped to tackle these issues not only for Europe. It also aspired to shape AI governance beyond its borders and “[to make] the Union a leader in the uptake of trustworthy AI” worldwide (Council of the European Union, 2024, 12). This ambition reflects the global clout that the EU has gathered over the decades, for example in environmental policy or data privacy (Bretherton & Vogler, 2006). Bradford (2012) famously identified a Brussels Effect through which the EU’s regulatory standards become de facto global norms as firms and governments in third countries adopt them, often voluntarily, to retain access to the EU’s market. Policymakers and scholars alike have hoped for similar EU influence in AI (Bradford, 2023; Dempsey et al., 2022; Krasodomski & Buchser, 2024; Meyers, 2023; Siegmann & Anderljung, 2022; Stix, 2022).
AI governance is a distinct policy field, however: the economic stakes are enormous, the technologies in question evolve rapidly, and the companies building them are concentrated in a handful of countries. Given these characteristics, we therefore treat the Brussels Effect in AI as an open question and ask: how do structural and contextual factors in the field of AI governance affect the likelihood and transformative potential of a Brussels Effect there? A transformative Brussels Effect would go beyond mere rule absorption, in which EU policies are transposed but without actual implementation or other forms of substantial impact (Börzel & Risse, 2003). The more EU rules change foreign practices “on the ground” by travelling there, the more transformative—and thus meaningful—the Brussels Effect is.
At the heart of this paper lies a topical puzzle: right from the first AIA draft in spring 2021, EU legislators had hoped that its AI rules would have global influence. A Brussels Effect in AI was part of the plan to compensate for Chinese and American technological leadership (Bradford, 2023) – not just a potential serendipitous side effect. It is against the background of this ambition that we inquire how structural and contextual features of AI governance, such as the high economic stakes involved, rapid technological change, concentrated corporate power, and regulatory uncertainty, have systematically shaped and, as we argue, limited EU influence beyond its borders.
AI governance is a young policy field, and the historical track record is limited. We cannot assess the strength of a Brussels Effect based on historical data alone—a clear handicap for our endeavor. This handicap is not coincidental but endemic to AI governance: it reflects a Collingridge-type dilemma, whereby policymakers are compelled to make regulatory commitments under conditions of considerable uncertainty about how the technology and its markets will evolve (Collingridge, 1980). The time at which regulatory choices remain relatively easy to revise is also the one at which their long-term domestic and global consequences are still hard to judge.
This dilemma shapes our own inquiry: if we want our contribution still to be part of an ongoing policy conversation, we cannot wait until the dust has settled and the policy field has completely stabilized. To our mind, the strength of a Brussels Effect in AI has implications other for other policy choices in the field: if the EU can exert global influence as a legislative front-runner, it can remain open to foreign AI companies, because rule setting elsewhere will eventually emulate the European approach. If, on the other hand, a Brussels Effect is systematically less pronounced in AI, the EU may need more hard-nosed policies to advance digital self-determination (Pohle & Thiel, 2020)—possibly including forms of digital industrial policy and stronger digital trade restrictions. As the stakes are high, we therefore find it imperative to establish what scholarship can reasonably add to tackling this question through a preliminary empirical examination.
Our analysis combines two inferential strategies. First, following an analogical mode of reasoning, we determine to which degree the preconditions for a successful Brussels Effect that have been observed in other policy fields are also present in AI. We take these preconditions from Bradford’s own work. We also, however, draw on the wider Europeanization literature (Börzel & Risse, 2003; Lavenex & Schimmelfennig, 2009; Risse et al., 2001; Schimmelfennig, 2010), of which the Brussels Effect is one specific version. Methodologically, our study resembles a most-similar comparison (Gerring, 2007) across policy fields, and the relevance of its approach extends beyond AI policy. Policy and technological developments have gathered pace, turning the fields scholars analyze into moving targets. In that respect, this study is an example of gauging the strength of a Brussels Effect under uncertainty, and the heuristic we offer has relevance also for other fields of technological development, both digital and not.
Second, empirics are limited but not absent, even as important AIA provisions have not entered into force yet and keep getting delayed through the Digital Omnibus that is being negotiated at the time of writing. Following the Brussels Effect conditions, we assess how independently or not the EU developed its own policy approach, what role independent standardization organizations play in AI governance, and to what degree other major AI powers have so far embraced EU rules. Although the ongoing AIA implementation makes a full ex-post evaluation impossible as of yet and our results therefore remain tentative, the policy dynamic thus far already allows specific inferences.
The “size” of a Brussels Effect is impossible to measure in the absence of reliable, operationalizable indicators. Whether it is “strong” or “weak” is in the eye of the beholder. Our contribution takes a different approach, positioning our findings relative to the optimistic initial expectations and ambitions. The analyses generates four Brussels Effect-skeptical arguments for AI. First, although the EU’s large market size would, in principle, favor a Brussels Effect in AI, informational disadvantages, its reliance on external standard-setters, enforcement delays, and the ability of major AI companies to customize models by jurisdiction mean that its regulatory capacity and leverage are limited.
Second, EU AI rules have been designed in the shadow of a potential competitive threat and thus fell short of potential aspirations from the beginning. Much EU AI governance has been geared towards a competitive AI race. Politicians have frequently rejected tighter rules because they would dent the EU’s AI competitiveness (Mügge, 2024; Paul, 2023). The global context shaped EU rules more than the other way around. That continues to be the case.
Third, what EU rule diffusion we have seen is superficial: general guiding principles—such as ethical frameworks or a risk-based approach—are easy to agree on and copy, given that what matters is the actual operationalization, implementation and enforcement of concrete rules. Also in the EU itself, many specific provisions are yet to be decided and will often draw on templates from outside the EU. The result is what we call “shallow harmonization”.
Fourth, with cutting-edge AI concentrated in the USA and China, the real litmus test for the EU’s global influence is its rules’ impact in those two jurisdictions. We remain skeptical—on both empirical and theoretical grounds—that the EU can or will have a significant independent impact where it would matter most, even if rule diffusion is more likely elsewhere.
Our analysis concentrates on the time between the European Commission’s first legislative proposal in 2021 and early 2025. We develop these arguments in two major steps. First, we review the Brussels Effect and the mostly sanguine assessments of its potential in AI. We then delve into the limits for such a dynamic more systematically, add additional insights borrowed from adjacent literatures, and apply those to AI, both conceptually and, where data is available, empirically.
2 EU Regulatory Ambitions in the Era of AI
The EU has earned a reputation as a regulatory superpower, and many Brussels policymakers have been optimistic that it will be one in AI governance, as well (Bradford, 2012; Bremmer, 2016; Li et al., 2023). The Commission, for example, argued that “[the] main ingredients are there for the EU to become a leader in the AI revolution” (European Commission, 2018a, 19). Three years later it dedicated an entire chapter to “Creating EU global leadership […]” in the field (European Commission, 2021, 7). Commission president Ursula von der Leyen dubbed the AIA a global “blueprint” (European Commission, 2023); the official EP statement celebrated the “world’s first comprehensive AI law” (European Parliament, 2023), which, argued the European Parliamentary Research Service, may “influence global practices of AI developers” (European Parliamentary Research Service, 2024, 5).
One key mechanism through which the EU has hoped to level the playing field and export its regulatory preferences is the Brussels Effect. In this regard, the General Data Protection Regulation (GDPR) has served as an example for the EU’s ability to set international standards (Li et al., 2023). Bradford had described the Brussels Effect as the EU’s ability “to promulgate regulations that shape the global business environment, leading to a notable ‘Europeanization’ of many important aspects of global commerce” (Bradford, 2019, 15). Central to her argument: the EU need not push its rules proactively. Instead, foreign companies and governments promote and diffuse them to access the EU market while avoiding compliance with divergent rule sets. In the de facto Brussels Effect, multinational firms comply with particularly stringent rules across all their markets, because that is less costly than producing to different specifications. In a second step (the de jure effect), the multinationals then “lobby their domestic governments to adopt these same standards in an effort to level the playing field against their domestic, non-export-oriented competitors” (Bradford, 2012, 16). For the Brussels Effect to materialize, Bradford identifies five conditions: sufficient market size, regulatory capacity, a preference for stringent rules (and political will to enforce them), the inflexibility of the targets being regulated, and non-divisibility – the difficulty to apply different product rules to different markets (Bradford, 2019).
Over time, expectations for a Brussels Effect in AI have been mixed. Early on, optimism prevailed. In the AIA’s early days, the key to EU power was mainly seen in its market size and regulatory prowess (Brattberg et al., 2020), even if Europe already lagged the USA and China in AI development (Straus, 2021). Siegmann and Anderljung (2022, 29) estimated the EU to be “no lower than 15% of the global [AI] market” (in a similar direction, Dempsey et al., 2022). Bradford has marshalled the “nearly 450 million relatively wealthy [EU] consumers” (Bradford, 2023, 326) as a boost to an AI Brussels Effect. And with the EU forging ahead with hard AI law, third countries might thankfully copy its rules to avoid reinventing the wheel, especially for high-risk AI systems (Engler, 2023). “[By] virtue of spearheading ‘trustworthy AI’, the EU has occupied a position where it has been able to shape the discourse on global AI governance discourse early on” (Stix, 2022, 938).
Already in 2022 Gstrein surmised that “probably the most relevant achievement of the AIA to date might be its pioneering role [..]. This in itself has an impact on many areas of society, also beyond the borders of the EU” (Gstrein, 2022, 19). Veale et al. noted that “the European Union and its member states have quickly emerged as a key player in the global regulation of AI [..]” (Veale et al., 2023, 13). Leading European policy think tanks have been similarly optimistic, with Meyers from the Centre for European Reform noting that “the EU [.] will continue to enjoy the most influence on global technology regulations” (Meyers, 2023).Footnote 1 And Krasodomski and Buchser from Chatham House have attested the AIA a capacity to “help to lead global change”, even when they wondered whether the AIA would equal the GDPR in its global effect (Krasodomski & Buchser, 2024).
As time progressed, however, skeptical voices multiplied. Ylönen (2026) argued that the Brussels Effect does not provide the proper analytical tools to capture the influence wielded by both governing bodies (in this case, the EU) and, increasingly, by private actors, such as multinational enterprises (MNEs). Regulatory competition allows the latter to arbitrage regulatory gaps through strategic engagement in implementation and enforcement, undermining the uniform global standard-setting that the Brussels Effect implies (Ylönen, 2026). Similarly, Hine (2024) found that “the power of big tech is outstripping any ‘Brussels effect’ from the EU’s AI Act” (Hine, 2024) – especially given the struggles of European competitor companies (Engler, 2022) and hence political pressure to promote those through regulatory lenience (Allison & Rugova, 2025; Haeck & O’Regan, 2025).
3 Limits of the Brussels Effect and Rule Externalization in AI
The Brussels Effect is a specific form of EU rule externalization (Bendiek & Stuerzer, 2023; Christen et al., 2022; Scott, 2019). It has most famously been articulated by Anu Bradford, with the GDPR (cf. Farrell & Newman, 2019) as the archetypal example. Beyond the difficulties concerning actual measurement (Li et al., 2023), the broader literature on rule externalization has drawn out factors that may hamper rule externalization and therefore limit the Brussels Effect, too (Büthe & Mattli, 2011; Lavenex & Schimmelfennig, 2009; Moravcsik, 1998).
Given that AI policy is still evolving, we use a two-pronged inferential strategy to assess the Brussels Effect in AI: we marshal empirical evidence where available, and we reason by analogy to establish whether the preconditions for a transformative Brussels Effect in AI are present. First, we consider the factors highlighted by Bradford herself and then move to the additional ones gleaned from externalization literature. While the Brussels Effect in AI was never universally accepted even in EU policy circles, it held considerable sway as a guiding assumption, particularly around the time of the AIA’s adoption. As we gauge to what degree it has materialized since, and why the optimism was excessive, we place Bradford’s central thesis, applied to the AIA, at the forefront of our inquiry.
To be sure, beyond potential rule externalization via a Brussels Effect, the EU is also active in bilateral and multilateral AI governance initiatives, such as the Council of Europe (CoE) and the transatlantic Trade and Technology Council (TTC, Schmitt, 2022). The CoE has already adopted an agreement, even if in a major setback the EU failed to make the CoE rules mandatory for companies, as well (Bertuzzi, 2024; Hoch, 2026). Many other initiatives hope to craft international AI soft law. EU presence in these governance forums is an alternative, and potentially important, channel for EU influence. However, as this paper concentrates on rule externalization via a Brussels Effect, we ignore this influence channel here.
4 Traditional Drivers: Market Size, Regulatory Capacity and Non-Divisibility
Ex ante, the EU’s market size favors a Brussels Effect in AI. While political actors across the Brussels institutions have been eager to create a legal framework for AI, the EU’s regulatory capacity has been less self-evident. As elsewhere, EU policymakers have inevitably been at an informational disadvantage vis-à-vis AI companies (Schuett, 2023). It remains unclear how big specific AI risks are, and what effective measures are to tackle them (Nordström, 2022), especially for general-purpose AI models (Moës & Ryan 2023). In consequence, legislators effectively outsourced risk assessment and the filling in of legislative details to either companies themselves or standard-setting bodies formally outside the EU. As the difficulty of crafting tractable rules has become clearer and implementation deadlines became untenable, the Commission had to delay actual enforcement of AIA provisions to August 2026, more than two years after the regulation’s adoption.
The EU’s leverage over foreign companies shrinks once the latter decide that, rather than complying with local rules, they simply do not offer products or disable certain features on EU markets. Meta, for example, has first withheld its latest AI model (“Meta AI”) in the EU because of legal uncertainties (Weatherbed, 2024)Footnote 2; Apple has blocked AI-enabled real-time translation features for its newest AirPods in the EU (Haeck & Parry, 2025). The Brussels Effect, in which companies are forced to produce to European specifications, and then push for those same rules elsewhere, thus highly depends on the AI product in question. Large language models (LLMs) can easily be modified, for example by disallowing them to process particular prompts or by checking whether the output generated violates local rules (Engler, 2023). Companies may appreciate a single rule set across jurisdictions. But the ease of customization lowers incentives to follow EU rules everywhere. When we consider the common preconditions for a successful Brussels Effect, the picture is mixed at best.
5 Preference for Strict Rules vs. Pre-Emptive Concessions
The Brussels Effect presupposes that stringent rules are formed domestically first and are then diffused: the EU transmits its strict regulatory preferences to the rest of the world, so the idea, generating regulatory outcomes abroad that differ substantially from what would have transpired without the EU’s influence. Consequently, the more EU rules have themselves been shaped by external constraints, the less we see a genuine transfer of domestic preferences to the rest of the world, even when EU rules do travel (cf. Howell, 2004). To what degree does the EU’s AI regime actually reflect stringent domestic preferences, rather than a pre-emptive adjustment to external constraints?
AI technologies (AITs) are widely held up as decisive for future economic prosperity (Ulnicane, 2022). The Commission has repeatedly highlighted the need not to fall (further) behind other major AI powers (European Commission, 2018a; Seidl & Schmitz, 2024). Because of first mover advantages and high economies of scale in AI, governments are wary of ‘missing the boat’, spurring AI development where possible. And the enormous resources necessary to develop cutting-edge AI encourage joining forces with established players. These characteristics shape the regulatory politics they generate and thereby the scope for a Brussels Effect.
Two competing impulses have characterized Commission thinking from the beginning. On the one hand, a well-crafted regulatory regime could itself be a key selling point of “AI made in Europe” (European Commission, 2018b, 1, European Commission, 2020). At the same time, EU AI strategy acknowledges that Europe lags the USA and China in AI development and that any regulatory framework should therefore be “innovation-friendly” and avoid undue obstacles for AI businesses (Paul, 2024). AI development and deployment needed to be boosted through sundry Commission initiatives, with the AIA only providing the guardrails for this push (Mügge, 2024).
EU AIA debates already reflected foreign competition as negotiators tried to balance restrictions and permissiveness vis-à-vis AITs, most visibly in the 2023 trilogue negotiations between the EP, the Commission, and national governments. Several member states backpedaled on restrictions for general-purpose AI models—the most capable systems at the time—once credible EU challenger firms to US tech dominance emerged (Bertuzzi, 2023; Volpicelli, 2023). That fall, executives from Germany’s Aleph Alpha and France’s Mistral AI had publicly opposed the regulation of LLMs (Henshall, 2023). In October 2023, French, German, and Italian business and economic ministers agreed on their “commitment for an innovation-friendly and risk-based approach, reducing unnecessary administrative burdens on companies that would hinder Europe’s ability to innovate” (Ministry of Enterprises and Made in Italy, 2023). In November, with AIA negotiations all but finished, Germany, France and Italy published a so-called non-paper (effectively a discussion or position paper) in which they pressed only for codes of conduct for LLMs, rather than prescriptive regulation, as the draft AIA had foreseen.
The EP remained steadfast in its opposition to rule dilution; the eleventh-hour compromise applied only light touch regulation to the LLMs currently developed in Europe. EU AIA rules had, in other words, themselves been watered down in the face of US competition. This pre-emptive adjustment blunted the EU’s regulatory ambition, and thus the export of genuinely European regulatory preferences. When those diluted rules then become global standards, they reflect relative EU impotence rather than its power (cf. Geradin & McCahery, 2004).
6 Enforceability: General Rules, Shallow Harmonization
A Brussels Effect can look impressive but be superficial. Are similarities between a country’s formal rules and European ones enough? Or does a Brussels Effect only exist once EU-like legislation has produced effects like those inside the EU? Here, the literature on Europeanization – of which the Brussels Effect arguably is one specific form – is useful. While this literature initially focused on member states and “the domestic impact of, and adaptation to, European governance” in them (Schimmelfennig, 2010, 319), it ventured outward from there and investigated domestic change in non-EU countries due to the EU’s external governance (Risse et al., 2001). Börzel and Risse (2003) for example distinguished between rule absorption, when EU policies are incorporated into domestic ones without substantive policy modification; accommodation, where formal policies do change, but not their essence or rationales; and transformation, so a genuine reorientation of domestic policies. Absorption without accommodation or transformation translates into a shallow Brussels Effect, especially when EU rules are rather superficial and can be implemented by third countries without significant adoption costs or efforts (Jørgensen et al., 2011).
AI has specific features that stand in the way of detailed rules with real teeth, limiting a Brussels Effect in this field. Because AITs evolve quickly, they require an adaptable legal framework (Nordström, 2022)—for example in the thresholds for systemically risky general-purpose AI (Moës & Ryan 2023). AIT governance inevitably is a permanent regulatory construction site. For that reason, jurisdictions hesitate to adopt rigid regulation—a dynamic that in turn limits wholesale rule export or import. And where rules overlap now, that does not imply congruence in the future. It is not a coincidence that the most prominent international AI agreement, the 2024 Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law of the Council of Europe, is so vague that it can hardly be enforced.
During much of the 2010s, “AI ethics” dominated regulatory debates (Jobin et al., 2019; Mittelstadt et al., 2016). Their thrust has been to identify normatively contentious real-world implications of AI, and then to articulate the principles for AI governance: the need for accountability, explainability, trustworthiness, non-discrimination, transparency, privacy, fairness, and so on (cf. Gasser & Mayer-Schönberger, 2024). Most of these principles are easy to agree on. Cross-border support for them is a far cry from regulatory harmonization, however. They sound appealing in the abstract but lack concreteness (Stamboliev & Christiaens, 2024) and remain too broad to offer tangible guidance (Whittlestone et al., 2019).
In a typical example, the AIA mandates in Article 10(2), (f) and (g) that
-
2.
Training, validation and testing data sets shall be subject to data governance and management practices appropriate for the intended purpose of the high-risk AI system. Those practices shall concern in particular:
-
(f);
examination in view of possible biases that are likely to affect the health and safety of persons, have a negative impact on fundamental rights or lead to discrimination prohibited under Union law, especially where data outputs influence inputs for future operations;
-
(g)
appropriate measures to detect, prevent and mitigate possible biases identified according to point (f). (AIA, p57, emphasis added)
-
(f);
The AIA does not specify, however, how the all-important qualifiers “appropriate”, “possible” and “likely” are defined. Without such clarity, on paper agreement could mask major differences across regimes.
The AIA will become legally tractable only through delegated acts, in which the Commission-based AI Office fills in details, or through technical AI standards, for example from the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) (Hennessy & Mügge, 2026). Technical standardization for the AIA would mainly run through the European Committee for Standardization (CEN) and the European Committee for Electrotechnical Standardization (CENELEC), as well as ISO/IEC. Since the AIA explicitly requires those technical standards, it is in these implementing measures that the real harmonization would happen—or not (Veale & Borgesius, 2021). Whether shared principles translate into equivalent regimes is therefore an open question.
In practice, it is often left to developers to interpret what “ethical AI” means, and disagreements may only be settled by the Court of Justice of the European Union (CJEU) as the Union’s highest legal authority (Hoofnagle et al., 2019, 71). If case law fills the holes in legal frameworks, however, local legal traditions and relevant precedents push similar-sounding principles in different directions.
Generative AI poses particular challenges. Rules for it had been added to the AIA in the later stages of the negotiations. The AIA requires providers of general-purpose systems to supply
information on the data used for training, testing and validation, where applicable, including type and provenance of data and curation methodologies (e.g. cleaning, filtering etc), the number of data points, their scope and main characteristics; how the data was obtained and selected as well as all other measures to detect the unsuitability of data sources and methods to detect identifiable biases, where applicable. (Council of the European Union, 2024, 269)
That sounds impressive. But in fact, it will be hard to review the enormous amounts of data thus supplied, let alone mitigate all potential risks (Ebers, 2025; Pouget & Zuhdi, 2024). That dynamic is aggravated once compliance is off-loaded to downstream users of AI systems, which are even harder to monitor than the relatively small number of companies that develop the technologies themselves.
The reliance on open-ended provisions or secondary implementing rules is not unique to AI governance. Nevertheless, it seriously limits a Brussels Effect in this field: measured against the ambitions of the “regulatory power Europe” narrative, a preference for stringent rules and the putative will to enforce them, such structural indeterminacy curtails the EU’s capacity to project its standards globally, even if it is not an exceptional feature of AI regulation as such.
Taken together, third country copying of general European frameworks, or international agreements that bear a heavy European imprint, does not equate to regulatory alignment. The AIA frequently lacks the detailed rules that make it legally tractable, and at least some of them will be crafted outside of the EU proper. In a sign how thorny hashing out the details is, technical standard setters CEN/CENELC have also postponed the expected publication of the standards that are to fill the AIA with life. Even where general principles of AI governance—such as a risk-based approach with different layers—are emulated elsewhere, that does not constitute a wholesale embrace of an EU-crafted regulatory regime.
7 Limited Diffusion
Finally, the relevance of a Brussels Effect not only varies with the degree to which it actually produces regulatory alignment, but also with the countries in which we observe such alignment. Rules are easy (and inconsequential) to adopt when they have few implications in a particular jurisdiction because there are few targets to which to apply them. It matters whether rules target products traded across borders or offer direct protections for citizens. If citizen protections are the issue, every additional country embracing, say, privacy protection rules, counts as a win – even if it has no big domestic AI companies. If, in contrast, rules target the production of goods and services, they may only matter in the limited number of countries where such production takes place – the major AI powers.
From a European perspective there are different reasons to care about a Brussels Effect. Where collective action problems exist, for example in environmental protection, emulation of stringent EU rules elsewhere is useful because that, too, contributes to solving genuinely global problems. Adoption of EU rules abroad also matters because it may pre-empt regulatory races to the bottom. And diffusion of rules seen to benefit individuals—for example by protecting their personal rights—may be valuable in and of itself, irrespective of any direct or indirect pay-off for Europe.
Applied to AI governance, some of these effects are much more pertinent than others. AI, too, generates thorny global collective action problems. These concern the development of militarily dangerous or undesirable AI applications (Bode & Huelss, 2023), the diffusion of dangerous AI-powered tools around the world, and the management of existential risks that the most powerful AI systems may constitute.
If the challenge is to prevent the development and spread of dangerously advanced AI systems, the countries that really matter are those that actually have advanced AI capabilities. Just as land-locked countries will matter little to ocean overfishing, nations outside the AI vanguard will not be decisive in curtailing harmful “AI races”. Inversely, just as countries without a fishing fleet can easily commit to stringent marine conservation rules, countries without an advanced tech sector need not hesitate to sign tough safety rules for frontier models—they do not build any to begin with.
For an AI Brussels Effect as a corrective to collective action problems, the actions of other major AI powers are key, above all USA and China, as well as of countries like the UK, Canada, Russia, or Japan, rather than those of countries that develop little AI themselves. But the intense competition and oligopolistic tendencies in advanced AI mean that both countries and companies developing it face strong incentives against rules that encumber technological advancement. As long as it remains relatively easy to customize digital products to the specificities of different markets, there is no reason to expect that leading AI companies would act as transmission belts for EU rules.
While the limited diffusion of EU rules and their watering-down through pre-emptive concessions in the first place present two related mechanisms, they are analytically distinct and entail more than a simple story of “who bows to whom” in global AI regulation: the EU’s willingness to make pre‑emptive concessions reflects internal dynamics whereby its regulatory ambition is curtailed ex ante in light of anticipated competitiveness and geopolitical concerns; by contrast, the limited receptiveness of other major AI powers, notably the US and China, to EU regulatory preferences points to constraints of EU rule externalization in AI in jurisdictions where this would matter most. While related, preemptive concessions and limited diffusion thus remain analytically distinct in their timing within global regulatory dynamics. The European AI landscape is clearly dominated by US companies. EU rules travelling across the Atlantic would have more impact than those diffusing anywhere else. The empirical record so far is disappointing. In October 2023, before the EU finalized AIA negotiations, the Biden administration issued an “Executive Order (EO) on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” (The White House, 2023). In abstract terms, the EO echoed many AIA themes (Bassini, n.d.), and it committed to “safe, responsible, fair, privacy-protecting, and trustworthy AI systems”, because the latter might “exacerbate societal harms such as fraud, discrimination, bias, and disinformation; displace and disempower workers; stifle competition; and pose risks to national security” (The White House, 2023).
That was not to last, however. President Trump rescinded the EO on his first day in office in January 2025, promising AI for military purposes and cementing US AI leadership with yet fewer strings attached (Thornhill, 2025). And a July 2025 Executive Order explicitly demanded that
[the] United States must not only lead in developing general-purpose and frontier AI capabilities, but also ensure that American AI technologies, standards, and governance models are adopted worldwide to strengthen relationships with our allies and secure our continued technological dominance. (Trump, 2025 emphasis added)
Where an AI Brussels Effect would have mattered most—in the USA—it was never likely, and it certainly is not under President Trump.
Other leading AI nations show a similar picture. Shaped by a different normative and ideological approach to social control and national security (Christou et al., 2025), China has little appetite to copy European provisions. China adopted its first AI rules in 2017 already, long before the Commission presented even a draft of its AIA, and the Chinese government has continued to set its own priorities since then (cf. Arcesati, 2021). When the Chinese DeepSeek released its LLM in January 2025, both the EU AI Office and several EU countries started investigations and Italy banned it altogether (Caroli, 2025). That said, the AI Office itself still had to fine-tune its classification guidelines for a model like DeepSeek’s R1 (García-Herrero & Krystyanczuk, 2025), not least because it is available open access. If DeepSeek were to sell versions of its models on the European market, it would have to comply with AIA provisions. But even then, it is improbable that it would skirt Chinese rules to attune its whole business model to European ones, given both the political and commercial importance of the Chinese market and political context. Legal emulation and ideational influence suggesting that China regards the European model as a benchmark worth emulating seems particularly unconvincing, considering the stark divergence between the two jurisdictions and China’s insistence on digital sovereignty (W. Li & Chen, 2024). Currently, we are not aware of any indications that Chinese firms act as transmission belts for European policies.
The United Kingdom under Rishi Sunak had ruled out a comprehensive regulatory framework for AI in the short term, instead touting voluntary agreements with companies to secure a competitive advantage for them. The Labour-led government assuming power in July 2024 announced plans to introduce a binding AI bill; instead of copying EU provisions, however, the intention has been to use rules different to those on the Continent to boost domestic firms’ competitiveness (Gross & Mnyanda, 2024). Canada published its Pan-Canadian AI Strategy in 2017, again ahead of the EU’s draft AIA. In June 2022, the government introduced the Artificial Intelligence and Data Act (AIDA) as part of its Digital Charter Implementation Act, which is under negotiation at the time of writing (Government of Canada, 2025). Also there, beyond the EU-Canadian agreement on a risk-based approach, nothing suggests an eagerness to import EU provisions.
The Japanese government published its Social Principles on Human-Centric AI in 2019, mirroring the OECD’s AI principles (Habuka, 2023). In abstract terms, they rhyme with the EU approach. But Japan has committed to a “sector-specific and soft-law-based approach”, rather than binding regulatory frameworks (Habuka, 2023, 6).
Finally, Russia has underlined its ambitions in the AI military space; its first major governmental initiative was a 10-point statement by the Ministry of Defense, laying out a military-specific development plan for AI (Petrella et al., 2021). Although its 2019 National Strategy for AI seems to focus on civilian matters, Russia has often highlighted its interest in upgrading its military equipment, including its opposition to banning the use of lethal autonomous weapon systems (Markotin & Chernenko, 2020; Nocetti, 2020). Even when we disregard the current geopolitical tension between Russia and the EU, this suggests a strongly distinct focus of and visions for AITs and offers little room for adoption of EU rules.
We cannot survey a potential AI Brussels Effect across all jurisdictions. A major reason to hope for one in the first place, however, would be that EU rules would spread to countries that play a significant role in AI development and deployment—above all the USA and China. We find some regulatory parallels, especially in a multi-tier risk-based approach (Christou et al., 2025). But there is no sign that these were consciously copied from the EU, rather than developed alongside them, or diffused through international forums such as the OECD. It remains conceivable that other countries would copy the AIA in a more wholesale fashion such that compliance with EU rules would automatically entail compliance there, as well—a key mechanism of the Brussels Effect (Bologa, 2025). But given their limited sway in AI, such diffusion would do little to address the collective action problems that would have EU policymakers hope for a Brussels Effect in AI in the first place.
8 Conclusion
In this article, we have outlined why we are skeptical about the scope for a Brussels Effect in AI. Legal frameworks around the world are still evolving. We have therefore gone beyond leveraging the limited empirical base we have so far. Using both the original Brussels Effect literature and adjacent strands, we have highlighted policy field features that promote or obstruct a Brussels Effect in general. In a second step, we have then established to what degree we find those in AI.
There is no absolute yardstick against which we could hold these findings; establishing, let alone quantitatively specifying, what a “likely” or “unlikely” Brussels Effect would be remains elusive.
However, in our analysis we have gauged the extent to which a Brussels Effect holds up in the empirical case of the EU’s AIA. Our findings suggest reasons to be skeptical: AI has specific characteristics that bear heavily on the dynamics of regulatory interdependence and diffusion. These include AITs’ heavy commercial implications and economic stakes, uncertainty about future applications and implications, the speed of its development, the substantial investment that AITs need, and their dual use character.
Our assessment has centered on four arguments: first, while the EU’s market size should theoretically support a Brussels Effect in AI, its influence is undermined by information asymmetries, dependence on external standard-setting bodies, slow enforcement, and the capacity of major AI firms to tailor their models to different regulatory environments.
Second, EU legislators crafted the AIA with a clear sense that the EU is locked into a competitive AI race of sorts (critically Bryson & Malikova, 2021; Lee, 2018). Some rules had explicitly been watered down to give EU AI firms a leg up. This accommodation of external pressure limits the degree to which a Brussels Effect could meaningfully export regulatory preferences.
Third, the real-world relevance of any observed Brussels Effect depends on the degree to which it changes the regulatory status quo in the target country. Departing from Bradford’s framework, it is easy to mistake third countries’ copying of rules from elsewhere for actual regulatory alignment when the rules themselves are vague, and when downstream operationalization and implementation may still diverge. Mock-harmonization is the result. By the same token, a Brussels Effect becomes less meaningful as enforcement becomes patchy, creating a chasm between on-paper rules and the alignment of in-practice provisions.
Finally, with cutting-edge AI currently concentrated in a few major jurisdictions, such as the USA and China, the real litmus test for the EU’s global influence is its rules’ impact in those jurisdictions. Diffusion of European AI rules to countries that have no internationally important AI sector may be beneficial for local populations there from a normative perspective. But detrimental regulatory races to the bottom can only be prevented if the jurisdictions with the highest global impact follow the EU’s lead. There, a Brussels Effect is neither visible nor, we argue, likely.
The potential of the AIA’s rule export is not only an academic matter. The AIA represents a pivotal test of the Brussels Effect precisely because it combines high regulatory ambition with a set of structural and geopolitical constraints that Bradford’s original account might not fully anticipate. Policymakers have had high hopes that their rules exert cross-border influence to internationalize EU preferences in AI to prevent a race to the bottom; in contrast, if other jurisdictions go their own way in rule setting more than had been anticipated, the consequences for EU AI policy would be significant. It could no longer, for example, trust that AI development elsewhere would follow normative EU imperatives. Notwithstanding agreement on some general principles, de facto divergence in AI rule setting would strengthen the case for a more robust EU AI policy, including support for a genuinely homegrown AI sector. In that sense, the limits of an AI Brussels Effect we have diagnosed in this article are also a wake-up call for crafting a more robust and independent European AI policy tout court.
Data Availability
Not applicable.
Notes
Both the Centre for European Reform and Chatham House were listed among the top 100 think tanks worldwide according to the 2020 Global Go To Think Tanks Index published by the Lauder Institute of the University of Pennsylvania (McGann 2021).
Although the European Commission has still not granted full clearance, Meta still rolled out the model in early 2025 (Kroet, 2025).
References
Acemoğlu, D., & Johnson, S. (2023). Power and progress. Our thousand-year struggle over technology and prosperity. Basic Books.
Allison, K., & Rugova, V. (2025). Is the brussels effect fading? April 8, 2026. https://cepa.org/article/is-the-brussels-effect-fading/
Arcesati, R. (2021). Lofty principles, conflicting interests: AI ethics and governance in China MERICS China Monitor: Leading actors driving each development: Government or government-affiliated academia, industry, civil society.
Bassini, M. The global race to regulate AI: Biden’s executive order spillover effects on the EU AI Act. Institute for European Policymaking. https://iep.unibocconi.eu/publications/global-race-regulate-ai-bidens-executive-order-spillover-effects-eu-ai-act
Bendiek, A., & Stuerzer, I. (2023). The Brussels Effect, European regulatory power and political capital: Evidence for mutually reinforcing internal and external dimensions of the Brussels Effect from the European digital policy debate. Digital Society, 2(1), 5. https://doi.org/10.1007/s44206-022-00031-1
Benjamin, R. (2019). Race after technology. Cambridge: Polity.
Bertuzzi, L. (2023). EU’s AI ambitions at risk as US pushes to water down international treaty. www.euractiv.com. June 4, 2024. https://www.euractiv.com/section/artificial-intelligence/news/eus-ai-ambitions-at-risk-as-us-push-to-water-down-international-treaty/
Bertuzzi, L. (2024). EU prepares to push back on private sector carve-out from international AI treaty. www.euractiv.com. June 6, 2024. https://www.euractiv.com/section/artificial-intelligence/news/eu-prepares-to-push-back-on-private-sector-carve-out-from-international-ai-treaty/
Börzel, T. A., & Risse, T. (2003). Conceptualizing the domestic impact of Europe. In K. Featherstone, & C. M. Radaelli (Eds.), The politics of europeanization (pp. 57–80). Oxford University Press. https://doi.org/10.1093/0199252092.003.0003
Bode, I., & Huelss, H. (2023). Constructing expertise: The front- and back-door regulation of AI’s military applications in the European Union. Journal of European Public Policy, 30(7), 1230–1254. https://doi.org/10.1080/13501763.2023.2174169
Bologa, A. (2025). Burying the brussels effect? AI Act inspires few copycats. CEPA. October 3, 2025. https://cepa.org/article/burying-the-brussels-effect-ai-act-inspires-few-copycats/
Bradford, A. (2012). The brussels effect. Northwestern University Law Review, 107(1), 1–68. https://scholarship.law.columbia.edu/faculty_scholarship
Bradford, A. (2019). The Brussels Effect: How the European Union rules the world. Oxford University Press USA-OSO.
Brattberg, E., Csernatoni, R., & Rugova, V. (2020). Europe and AI: Leading, lagging behind, or carving its own way?.
Bradford, A. (2023). Digital empires: The global battle to regulate technology. Oxford University Press.
Bremmer, I. (2016). What happens next with Europe’s new regulation of big tech. Council on foreign relations. https://www.cfr.org/blog/global-agenda-navigating-contested-global-spaces
Bretherton, C., & Vogler, J. (2006). The European Union as a global actor (2nd ed.). Routledge.
Bryson, J. J., & Malikova, H. (2021). Is there an AI cold war? Global Perspectives, 2(1). https://doi.org/10.1525/gp.2021.24803
Büthe, T., & Mattli, W. (2011). The new global rulers: The privatization of regulation in the world economy. Princeton University Press. https://doi.org/10.1515/9781400838790
Caroli, L. (2025). DeepSeek: A problem or an opportunity for Europe? Center for Strategic & International Studies. March 19, 2025. https://www.csis.org/analysis/deepseek-problem-or-opportunity-europe
Christen, E., Meyer, B., Oberhofer, H., Hinz, J., Kamin, K., & Wanner, J. (2022). The Brussels Effect 2.0: How the EU sets global standards with its trade policy. Austrian Institute of Economic Research.
Christou, G., Meyer, T., & Fanni, R. (2025). The European union: Assessing global leadership through actorness in artificial intelligence. Journal of European Integration, 47(3), 383–401. https://doi.org/10.1080/07036337.2024.2377200
Collingridge, D. (1980). The social control of technology. London New York: Frances Pinter St. Martin’s press.
Council of the European Union. (2024). Proposal for a regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain Union legislative acts - analysis of the final compromise text with a view to agreement.
de Vries, A. (2023). The growing energy footprint of artificial intelligence. Joule, 7(10), 2191–2194. https://doi.org/10.1016/j.joule.2023.09.004
Dempsey, M., McBride, K., Haataja, M., & Bryson, J. J. (2022). Transnational digital governance and its impact on artificial intelligence. In J. B. Bullock, Y.-C. Chen, J. Himmelreich, V. M. Hudson, A. Korinek, M. M. Young, & B. Zhang (Eds.), The Oxford handbook of AI governance (pp. 253–275). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780197579329.013.16
Ebers, M. (2025). Truly risk-based regulation of artificial intelligence how to implement the EU’s AI act. European Journal of Risk Regulation, 16(2), 684–703. https://doi.org/10.1017/err.2024.78
Engler, A. (2022). The EU AI act will have global impact, but a limited brussels effect. https://www.brookings.edu/articles/the-eu-ai-act-will-have-global-impact-but-a-limited-brussels-effect/
Engler, A. (2023). The EU and U.S. diverge on AI regulation: A transatlantic comparison and steps to alignment. https://www.brookings.edu/articles/the-eu-and-us-diverge-on-ai-regulation-a-transatlantic-comparison-and-steps-to-alignment/#anchor3
European Commission, & Ursula. (2023). State of the Union address by President von Der Leyen. European Commission. April 16, 2024. https://ec.europa.eu/commission/presscorner/detail/en/speech_23_4426
European Commission. (2018a). Communication from the Commission to the European Parliament, the European Council, the Council, the European Economic and Social Committee and the Committee of the Regions: Artificial intelligence for Europe [COM(2018) 237 final].
European Commission. (2018b). Communication from the Commission to the European Parliament, the European Council, the Council, the European Economic and Social Committee and the Committee of the Regions: Coordinated plan on artificial intelligence [COM(2018) 795 final]
European Commission. (2020). White paper on artificial intelligence: A European approach to excellence and trust [COM(2020) 65 final]. https://ec.europa.eu/commission/sites/beta-political/files/political-guidelines-next-commission_en.pdf
European Commission. (2021). Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions: Fostering a European approach to artificial intelligence [COM(2021) 205 final].
European Parliament. (2023). EU AI Act: First regulation on artificial intelligence. https://www.europarl.europa.eu/news/en/headlines/society/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence
European Parliamentary Research Service. (2024). The global reach of the EU’s approach to digital transformation.
Farrell, H., & Newman, A. L. (2019). Weaponized interdependence: How global economic networks shape state coercion. International Security, 44(1), 42–79. https://doi.org/10.1162/isec_a_00351
García-Herrero, A., & Krystyanczuk, M. (2025). The geopolitics of artificial intelligence after DeepSeek. Bruegel | The Brussels-based economic think tank. March 19, 2025. https://www.bruegel.org/first-glance/geopolitics-artificial-intelligence-after-deepseek
Gasser, U., & Mayer-Schönberger, V. (2024). Guardrails: Guiding human decisions in the age of AI. Princeton University Press.
Geradin, D., & McCahery, J. A. (2004). Regulatory co-operation: Transcending the regulatory competition debate. In J. Jordana, & D. Levi-Faur (Eds.), The politics of regulation. Edward Elgar Publishing. https://doi.org/10.4337/9781845420673.00014
Gerring, J. (2007). Case study research: Principles and practices. In Case study research: Principles and practices (pp. 86–150). Cambridge University Press. https://doi.org/10.1017/CBO9780511803123.008
Government of Canada. (2025). The Artificial Intelligence and Data Act (AIDA) - Companion document. May 26, 2025. https://ised-isde.canada.ca/site/innovation-better-canada/en/artificial-intelligence-and-data-act-aida-companion-document
Gross, A., & Mnyanda, L. (2024). Starmer plans to introduce AI bill in King’s Speech. Financial Times. January 6, 2025. https://www.ft.com/content/1013c46f-247b-4d47-8e0f-ab7387b4f22c
Gstrein, O. J. (2022). European AI regulation: Brussels effect versus human dignity? Zeitschrift für Europarechtliche Studien, 25(4), 755–772. https://doi.org/10.5771/1435-439X-2022-4-755
Habuka, H. (2023). Japan’s approach to AI regulation and its impact on the 2023 G7 presidency.
Haeck, P., & O’Regan, E. (2025). Brussels is done being the World’s digital policeman. POLITICO. April 8, 2026. https://www.politico.eu/article/brussels-police-world-digital-tech-us-china-regulations/
Haeck, P., & Parry, J. (2025). Apple blocks translation AirPods in EU over regulatory concerns. POLITICO. October 3, 2025. https://www.politico.eu/article/apple-blocks-airpods-translation-eu-because-regulation/
Hennessy, O., & Mügge, D. (2026). European standardization in flux: navigating delegation and control in AI governance. Journal of Standardization, 5, OnlineFirst, https://doi.org/10.59490/jos.2026.8346
Henshall, W. (2023). E.U.’s AI regulation could be softened after pushback. TIME. July 17, 2024. https://time.com/6338602/eu-ai-regulation-foundation-models/
Hine, E. (2024). Artificial intelligence laws in the US states are feeling the weight of corporate lobbying. Nature, 633(8030), S15–S15. https://doi.org/10.1038/d41586-024-02988-0
Hoch, S. (2026). Shaping global AI governance? assessing EU actorness in the negotiations of the council of europe’s AI framework convention. JCMS: Journal of Common Market Studies, 1–19 https://doi.org/10.1111/jcms.70151
Hoofnagle, C. J., Sloot, B. V. D., & Borgesius, F. Z. (2019). The European Union general data protection regulation: What it is and what it means. Information & Communications Technology Law, 28(1), 65–98. https://doi.org/10.1080/13600834.2019.1573501
Howell, K. E. (2004). Uploading, downloading and European integration: Assessing the Europeanisation of UK financial services regulation. Journal of Banking Regulation, 6(1), 53–68. https://doi.org/10.1057/palgrave.jbr.2340181
Jørgensen, K. E., Oberthür, S., & Shahin, J. (2011). Introduction: Assessing the EU’s performance in international institutions – conceptual framework and core findings. Journal of European Integration, 33(6), 599–620. https://doi.org/10.1080/07036337.2011.606681
Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389–399. https://doi.org/10.1038/s42256-019-0088-2
Krasodomski, A., & Buchser, M. (2024). The EU’s new AI Act could have global impact. April 15, 2024. https://www.chathamhouse.org/2024/03/eus-new-ai-act-could-have-global-impact
Kroet (2025). Meta’s AI tool still not cleared for use by EU commission. euronews. October 3, 2025. http://www.euronews.com/next/2025/03/24/metas-ai-tool-still-not-cleared-for-use-by-eu-commission
Lavenex, S., & Schimmelfennig, F. (2009). EU rules beyond EU borders: Theorizing external governance in European politics. Journal of European Public Policy, 16(6), 791–812. https://doi.org/10.1080/13501760903087696
Lee, K.-F. (2018). AI superpowers. China, Silicon Valley, and the new world order. Houghton Mifflin.
Li, W., & Chen, J. (2024). From Brussels effect to gravity assists: Understanding the evolution of the GDPR-inspired personal information protection law in China. Computer Law & Security Review, 54, 105994. https://doi.org/10.1016/j.clsr.2024.105994
Li, S., Schütte, B., & Sankari, S. (2023). The ongoing AI-regulation debate in the EU and its influence on the emerging economies: A New case for the “Brussels Effect”? In M. Findlay, L. M. Ong, & W. Zhang (Eds.), Elgar companion to regulating AI and big data in emerging economies (pp. 22–41). Edward Elgar Publishing. https://doi.org/10.4337/9781785362408.00007
Markotin, N., & Chernenko, E. (2020). Developing artificial intelligence in Russia: Objectives and reality. Carnegie. https://carnegiemoscow.org/commentary/82422
McGann, J. G. (2021). 2020 Global Go To Think Tank Index Report. The Lauder Institute, University of Pennsylvania https://repository.upenn.edu/entities/publication/9f1730fa-da55-40bd-a1f4-1c2b2346b753
Meyers, Z. (2023). In tech, the death of the brussels effect is greatly exaggerated. Centre for European Reform. April 15, 2024. https://www.cer.eu/insights/tech-death-brussels-effect-greatly-exaggerated
Ministry of Enterprises and Made in Italy. (2023). Italy, Germany and France agree on strengthening their cooperation on artificial intelligence. Ministry of enterprises and made in Italy. July 17, 2024. https://www.mimit.gov.it/en/media-tools/news/italy-germany-and-france-agree-on-strengthening-their-cooperation-on-artificial-intelligence
Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016). The ethics of algorithms: Mapping the debate. Big Data & Society, 3(2), 2053951716679679. https://doi.org/10.1177/2053951716679679
Moës, N., & Ryan, F. (2023). Heavy is the Head that Wears the Crown: A risk-based tiered approach to governing General Purpose AI. The Future Society.
Moravcsik, A. (1998). The choice for Europe: Social purpose and state power from Messina to Maastricht. 1. publ., 1. print. Cornell Univ. Press.
Mügge, D. (2024). EU AI sovereignty: For whom, to what end, and to whose benefit? Journal of European Public Policy, 31(8), 2200–2225. https://doi.org/10.1080/13501763.2024.2318475
Nocetti, J. (2020). The outsider: Russia in the race for artificial intelligence.
Nordström, M. (2022). AI under great uncertainty: Implications and decision strategies for public policy. AI & Society, 37(4), 1703–1714. https://doi.org/10.1007/s00146-021-01263-4
Paul, R. (2023). The politics of regulating artificial intelligence technologies: A competition state perspective. In R. Paul, E. Carmel, & J. Cobbe (Eds.), Handbook on public policy and artificial intelligence. https://ssrn.com/abstract=4272867
Paul, R. (2024). European artificial intelligence “trusted throughout the world”: Risk-based regulation and the fashioning of a competitive common AI market. Regulation and Governance, 18(4), 1065–1082. https://doi.org/10.1111/rego.12563
Petrella, S., Miller, C., & Cooper, B. (2021). Russia’s artificial intelligence strategy: The role of state-owned firms. Orbis, 65(1), 75–100. https://doi.org/10.1016/j.orbis.2020.11.005
Pohle, J., & Thiel, T. (2020). Digital sovereignty. Internet Policy Review, 9(4), 1–19. https://doi.org/10.14763/2020.4.1532
policy field has completely stabilized. ToHoch, S. (2026). Shaping Global AIGovernance? Assessing EU Actorness in the Negotiations of the Council of Europe’s AI Framework Convention. JCMS: Journal of Common MarketStudies, 1–19 https://doi.org/10.1111/jcms.70151
Pouget, H., & Zuhdi, R. (2024). AI and product safety standards under the EU AI Act. Carnegie. https://carnegieendowment.org/2024/03/05/ai-and-product-safety-standards-under-eu-ai-act-pub-91870
Risse, T., Cowles, M. G., & Caporaso, J. (2001). 1. Europeanization and domestic change: Introduction. In M. G. Cowles, J. Caporaso, & T. Risse (Eds.), Transforming Europe (pp. 1–20). Cornell University Press. https://doi.org/10.7591/9781501723575-002
Schimmelfennig, F. (2010). Europeanisation beyond the member states. (ZSE)/Journal for Comparative Government and European Policy, 8(3), 319–339. https://doi.org/10.5771/1610-7780-2010-3-319
Schmitt, L. (2022). Mapping global AI governance: A nascent regime in a fragmented landscape. AI and Ethics, 2(2), 303–314. https://doi.org/10.1007/s43681-021-00083-y
Schuett, J. (2023). Defining the scope of AI regulations. Law, Innovation and Technology, 15(1), 60–82. https://doi.org/10.1080/17579961.2023.2184135
Scott, J. (2019). The global reach of EU law. In C. Marise, & J. Scott (Eds.), EU law beyond EU borders: The extraterritorial reach of EU law (pp. 21–63). Oxford University Press. https://doi.org/10.1093/oso/9780198842170.003.0002
Seidl, T., & Schmitz, L. (2024). Moving on to not Fall behind? Technological sovereignty and the “geo-dirigiste” turn in EU industrial policy. Journal of European Public Policy, 31(8), 2147–2174. https://doi.org/10.1080/13501763.2023.2248204
Siegmann, C., & Anderljung, M. (2022). The brussels effect and artificial intelligence: How EU regulation will impact the global AI market.
Simons, J. (2023). Algorithms for the people: Democracy in the age of AI. Princeton University Press.
Stamboliev, E., & Christiaens, T. (2024). How empty is trustworthy AI? A discourse analysis of the ethics guidelines of trustworthy AI. Critical Policy Studies, 1–18. https://doi.org/10.1080/19460171.2024.2315431
Stix, C. (2022). The ghost of AI governance past, present, and future: AI governance in the European Union. In J. B. Bullock, Y.-C. Chen, J. Himmelreich, V. M. Hudson, A. Korinek, M. M. Young, & B. Zhang (Eds.), The Oxford handbook of AI governance. Oxford University Press. https://doi.org/10.1093/oxfordhb/9780197579329.001.0001
Straus, J. (2021). Artificial intelligence – challenges and chances for Europe. European Review, 29(1), 142–158. https://doi.org/10.1017/S1062798720001106
The White House. (2023). Executive order on the safe, secure, and trustworthy development and use of artificial intelligence. https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/
Thornhill, J. (2025). Trump is becoming the technoking of America. Financial times. January 24, 2025. https://www.ft.com/content/ffcd84d1-822c-492c-94e2-07785cfe3524
Trump, D. (2025). Promoting the export of the American technology stack. October 2, 2025. https://www.whitehouse.gov/presidential-actions/2025/07/promoting-the-export-of-the-american-ai-technology-stack/
Ulnicane, I. (2022). Emerging technology for economic competitiveness or societal challenges? Framing purpose in artificial intelligence policy. Global Public Policy and Governance, 2(3), 326–345. https://doi.org/10.1007/s43508-022-00049-8
Veale, M., & Borgesius, F. Z. (2021). Demystifying the draft EU artificial intelligence act — analysing the good, the bad, and the unclear elements of the proposed approach. Computer Law Review International, 22(4), 97–112. https://doi.org/10.9785/cri-2021-220402
Veale, M., Matus, K., & Gorwa, R. (2023). AI and global governance: Modalities, rationales, tensions. Annual Review of Law and Social Science, 19(1), 255–275. https://doi.org/10.1146/annurev-lawsocsci-020223-040749 www.annualreviews.org
Volpicelli, G. (2023). Power grab by France, Germany and Italy threatens to kill EU’s AI bill. POLITICO. July 17, 2024. https://www.politico.eu/article/france-germany-power-grab-kill-eu-blockbuster-ai-artificial-intelligence-bill/
Weatherbed, J. (2024). Meta Won’t release its multimodal Llama AI model in the EU. The Verge. April 3, 2025 https://www.theverge.com/2024/7/18/24201041/meta-multimodal-llama-ai-model-launch-eu-regulations
Whittlestone, J., Alexandrova, A., Nyrup, R., & Cave, S. (2019). The role and limits of principles in AI ethics: Towards a focus on tensions. In AIES, 2019 - Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society, Association for Computing Machinery, Inc (pp. 195–200). https://doi.org/10.1145/3306618.3314289
Ylönen, M. (2026). Reconceptualising the brussels effect. JCMS: Journal of Common Market Studies, 64(1), 100–124. https://doi.org/10.1111/jcms.13731
Acknowledgements
We gratefully acknowledge the support for this research received by the Vici grant of the Dutch Research Council (NWO; grant VI.C.211.032). We also want to thank our colleagues from the RegulAite project, as well as Matti Ylönen for providing valuable comments to earlier versions of this article.
Funding
This work was supported by the RegulAite project, funded by Dutch Research Council under the Vici grant (NWO; VI.C.211.032).
Author information
Authors and Affiliations
Contributions
All authors have accepted responsibility for the entire content of this manuscript and consented to its submission to the journal, reviewed all the results and approved the final version of the manuscript. S.H. and D.M. conceived the study; S.H. primarily conducted the analysis and wrote the initial draft; D.M. made significant revisions and amendments.
Corresponding author
Ethics declarations
Competing Interests
The authors do not have any competing interests to declare that are relevant to the content of this article.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
Hoch, S., Mügge, D. EU Influence in Global AI Governance and its Limits. Digit. Soc. 5, 45 (2026). https://doi.org/10.1007/s44206-026-00284-0
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1007/s44206-026-00284-0