The article opens with what was, in 2012, a heterodox claim. Europe in the years after the eurozone crisis was widely described as a power in decline: militarily weak, demographically stagnant, monetarily fragile, geopolitically irrelevant. Anu Bradford — then, as now, a professor of law at Columbia — argued that this consensus had missed something essential. Outside the conventional registers of power, the European Union had built an extraordinarily effective mechanism for projecting its preferences onto the rest of the world. It did so without armies, without sanctions, and without the cooperation of other states. It did so through the unilateral promulgation of regulations that the global market had reasons of its own to obey.1 The phenomenon already had a name in domestic American legal scholarship — the California Effect, theorised by David Vogel — but it had not been described systematically as a feature of international regulatory order. Bradford’s contribution was to provide that systematic account, to identify the precise conditions under which it operates, and to argue that the European Union, in the early twenty-first century, was the only jurisdiction in which all of those conditions converged.
The third pathway to regulatory convergence
The conceptual core of Bradford’s argument is a distinction. Most scholarship on international regulatory convergence treats convergence as the product of either cooperation or coercion. Bradford identifies a third pathway that operates between the two, and the rest of the article is dedicated to specifying it.
Political regulatory globalisation occurs when states converge on common standards through negotiated agreements: treaties, conventions, regulatory dialogues, harmonisation regimes administered by international institutions. The Kyoto Protocol is one canonical example; the Basel accords on banking regulation are another. Coercive regulatory globalisation occurs when one state imposes its standards on another through threats, sanctions, or conditional incentives. The United States has used this instrument extensively in areas such as counterterrorism and drug enforcement, conditioning aid, trade access, and diplomatic recognition on the adoption of preferred rules.2
Unilateral regulatory globalisation, the phenomenon Bradford names the Brussels Effect, is neither. It is the migration of one jurisdiction’s standards into another’s territory without the former actively imposing them or the latter willingly adopting them. The mechanism is the market itself. The regulating state passes rules that apply only to its own jurisdiction; foreign firms wishing to retain access to its market adjust their conduct or production to comply; and the architecture of modern global supply chains — the scale economies of uniform production, the technical impossibility of jurisdictionally segregating data or transactions, the legal indivisibility of certain corporate actions — propagates that adjustment to the firms’ worldwide operations. The strict regulator’s rule becomes the global rule. No treaty is signed. No coercion is applied. No consent is given by the affected states. The market has done the work that international institutions failed to do.
The intellectual antecedent is Vogel’s analysis of the California Effect — the observation that California, by virtue of its market size and preference for strict consumer and environmental standards, was able to set de facto national standards within the United States.3 Manufacturers seeking access to the California market complied with California rules; the scale economies of uniform production then propagated those rules to their entire output, even where other states maintained more permissive standards. Bradford’s analytic move was to recognise that the same logic operates at a global scale, that the European Union plays California’s role in the international system, and that the literature on regulatory globalisation had failed to account for this. The framework she develops is intended to identify the precise conditions that enable a jurisdiction to exercise this kind of unilateral influence, the mechanism by which it propagates, the political economy that drives the regulating jurisdiction to use it, and the structural limits that constrain it. The article advances a five-part theoretical scheme, applies that scheme to five extended case studies, examines the EU’s motivations both external and internal, considers when unilateralism is preferred to multilateral negotiation, and concludes by mapping the constraints — most of them internal — that determine where the Effect operates and where it does not.
The contribution to the international relations literature is sharp and explicit. Daniel Drezner had argued that great-power consensus produces regulatory convergence and great-power disagreement produces rival standards, with the outcome of any contest depending on each power’s ability to assemble coalitions of adopters.4 Beth Simmons had shown that, in some financial domains, lenient regulators voluntarily adopt strict standards in order to attract capital, producing a market-driven race to the top that benefits everyone.5 Bradford’s argument cuts across both. The Brussels Effect produces de facto convergence even amid great-power disagreement, because the outcome of the regulatory contest is predetermined by the structural logic of the situation: the stricter regulator prevails by virtue of being stricter. And the convergence it produces is not voluntary in Simmons’s sense. Foreign firms would generally prefer different rules. They comply because the opportunity costs of non-compliance are intolerable. The phenomenon is what Lloyd Gruber called “go-it-alone power” — the capacity of a dominant actor to proceed without the consent of others, and thereby to change the calculus of the others such that they emulate, however reluctantly, the dominant standard.6
The five conditions
Part I of the article advances a five-part theoretical framework. A jurisdiction can globalise its standards unilaterally if, and only if, all five conditions hold. The framework is the analytical heart of the paper, and each condition merits separate examination.
The five conditions are: market power, regulatory capacity, preference for strict rules, predisposition to regulate inelastic targets, and the nondivisibility of standards. Each is individually necessary; together they are jointly sufficient.7 The framework’s strength as an analytical tool comes from its precision: it specifies not only what produces the Brussels Effect, but also what its absence predicts. A jurisdiction lacking any one of the five cannot generate the Effect, regardless of how richly it satisfies the others. A regulated domain lacking nondivisibility will not be globalised, even when the regulating jurisdiction has the largest market, the strongest institutions, the strictest rules, and the most inelastic targets.
Market power is the precondition most discussed in prior literature and the one Bradford treats with the least originality. A jurisdiction must have a large enough internal market that exclusion from it is costly enough for foreign producers to absorb the adjustment costs of compliance. The relevant calculation, Bradford specifies, is not market size in absolute terms but the ratio of exports to the strict jurisdiction relative to sales in the firm’s home or third-country markets. The greater this ratio, and the lower the firm’s ability to divert trade to alternative markets, the stronger the incentive to adjust. The EU’s roughly five hundred million consumers, its GDP of nearly sixteen trillion dollars at the time of writing, and its status as the world’s largest importer of goods and services supplied this condition in abundance, particularly for affluent consumer goods where European demand could not be replicated elsewhere.8 Bradford notes that the United States, China, and Japan also possess markets large enough to satisfy this condition in principle. The constraint comes from the other four.
Regulatory capacity is the condition Bradford treats with the most originality, and it is the one that does the most work in narrowing the universe of potential Brussels Effect generators. Market power alone is inert without institutions capable of producing rules, enforcing them against domestic and foreign actors, and imposing sanctions for non-compliance. Bradford defines regulatory capacity as comprising both institutional structures (administrative agencies, sanction authority, technical expertise) and political delegation (the actual legal competence to act). She argues that this condition disqualifies most large markets. China’s antitrust law was enacted only in 2008, and its competition agencies are still building the regulatory expertise that the European Commission has accumulated since 1957 and the United States Department of Justice since 1890.9 The institutional development gap is measured in decades, not years. Among large markets, only the United States and the European Union genuinely satisfy this condition at the level required to externalise standards globally. Bradford emphasises that EU regulatory capacity is uneven across policy domains — most extensive in trade, competition, and single-market matters where supranational competence is well-established, most limited in areas like foreign and security policy where member states have retained authority.10 The Effect can only operate in domains where the EU has, by treaty, the competence to produce binding rules.
Preference for strict rules is the political condition. To be a global regulator, the state must subscribe to domestic standards that prevail over more lenient standards by virtue of being more demanding. Bradford traces a historical reversal that placed the EU rather than the United States in this role. Until the 1980s, U.S. consumer and environmental standards were the strictest in the OECD, and European firms adjusted upward to American rules in a transatlantic California Effect that mirrors today’s pattern in reverse.11 By the 2000s, the position had migrated to Brussels. The proximate causes she identifies are several: the European commitment to the precautionary principle, which authorises regulatory intervention in the presence of scientific uncertainty rather than requiring proof of unreasonable risk; the constitutional centrality of the social market economy to EU identity, formalised in the Lisbon Treaty’s Article 1; the lower ideological polarisation of European elites compared to their American counterparts, which makes them more responsive to public demand for stringent regulation; and the structural difference between an EU regulatory culture that emphasises the costs of inaction (the risks of “false negatives”) and an American regulatory culture that emphasises the costs of action (the risks of “false positives” and unnecessary intervention).12 Bradford is careful to acknowledge exceptions — the United States regulates smoking more strictly than the EU, for instance — but the dominant direction of regulatory divergence since the late 1980s has been Brussels-strict, Washington-permissive.
The predisposition to regulate inelastic targets narrows the regulatory space in which the Brussels Effect can operate. Strict standards globalise only if they regulate targets that cannot escape by moving. Consumers cannot relocate to circumvent product safety rules, food safety regulations, or pharmaceutical approval requirements; the affluent European consumer remains in the EU regardless of how stringent the regulations governing what may be sold to them. Capital, by contrast, is mobile. If the EU were to impose corporate tax harmonisation at high levels, firms could re-incorporate elsewhere. If it imposed a financial transactions tax, trading activity could and would divert to non-EU financial centres. Labour is partially elastic: firms employing workers in Europe must follow European labour rules, but can simultaneously employ workers in less protective jurisdictions under quite different terms. The EU’s focus on consumer markets, food safety, chemical regulation, and competition policy reflects this structural constraint, and explains the systematic pattern of where the Brussels Effect operates and where it does not. Domains where targets can move are not amenable to unilateral globalisation; domains where targets cannot are.13
Nondivisibility is the fifth condition and, in some respects, the most analytically distinctive contribution of the article. It is the mechanism through which a regulation that formally applies only inside the EU becomes a regulation that effectively applies everywhere. Bradford distinguishes three forms, and the distinction is essential because each form generates the Effect through a different concrete mechanism.
Legal nondivisibility occurs when the regulated transaction cannot be partitioned across jurisdictions. The paradigmatic case is the global merger. A merger between two firms cannot proceed in one jurisdiction and be blocked in another; either the merger consummates or it does not. The strictest competition regulator with jurisdiction over the transaction therefore controls the global outcome. When the European Commission blocked the General Electric–Honeywell merger in 2001 after the United States Department of Justice had cleared it, this was not because the Commission’s jurisdiction was greater than the DOJ’s; it was because the no-merger outcome dominated the merger outcome by the logic of indivisibility.
Technical nondivisibility occurs when the architecture of the regulated activity does not permit jurisdictional segregation. Data flows are the canonical example. The infrastructure of the internet, the architecture of corporate data systems, and the technical impossibility of reliably identifying which user is a European citizen mean that firms cannot easily maintain a European privacy regime distinct from their regime elsewhere. The cheapest and often the only feasible compliance strategy is to apply the European standard globally. Bradford anticipated, accurately, that this dynamic would only intensify as data-intensive industries grew, and the post-GDPR experience has confirmed it.
Economic nondivisibility occurs when scale economies make dual production uneconomical even where it is legally and technically feasible. A chemical manufacturer that has reformulated its products to meet REACH standards could in principle maintain a parallel non-REACH-compliant production line for non-EU markets, but the cost of operating two production processes — separate facilities, separate quality assurance, separate distribution chains — usually exceeds the cost of selling the reformulated product worldwide. The standard propagates not because the firm prefers it, but because the alternative is more expensive.
The three forms of nondivisibility are not mutually exclusive. The privacy case typically involves both technical and economic nondivisibility; the chemicals case involves economic nondivisibility reinforced by the supply-chain decisions of downstream users; the antitrust case is principally legal but can be reinforced by reputational and structural considerations. What unites them is that, in each, the firm’s rational private choice — to maintain a single standard — converts the EU’s regulation into a global one. The Effect operates through the firms, not against them.
Market forces are sufficient to create involuntary incentives to adjust to the rules of the strict regulator. In other words, unilateral regulatory globalization entails the dominant jurisdiction imposing an incentive to adjust, followed by reluctant emulation by market participants. — Anu Bradford
De facto and de jure cascades
The Brussels Effect operates in two registers, and the article is careful to distinguish them. The distinction has consequences both for how the empirical evidence of the phenomenon should be read and for how the literature on regulatory globalisation has previously underestimated its scope.
The de facto Brussels Effect is the primary mechanism. It does not require any formal action by another state. A foreign firm reformulates its products to meet REACH standards, amends its data architecture to satisfy European privacy rules, restructures its merger documentation to anticipate Commission scrutiny, or alters its supply chain to ensure chemical compliance — all in order to retain access to the EU market. Once that adjustment has been made, the firm — for reasons of scale economy, technical architecture, or legal nondivisibility — applies the new standard to its entire global operation. The EU rule has become the global rule. No legislator outside the EU has formally adopted it. No treaty has been signed. No coercion has occurred. The market has done the work that international law would otherwise have required.14
The de jure Brussels Effect is the secondary mechanism. It emerges when firms that have already adjusted to EU standards lobby their home governments to adopt those same standards. The political economy is direct: an exporting firm that has incurred the adjustment costs of REACH compliance has a strong interest in having its domestic competitors face equivalent costs, which they will not unless the home jurisdiction adopts equivalent rules. Civil society organisations frequently amplify this dynamic, treating the EU standard as a normative benchmark and pressing for its formal incorporation into domestic legislation. The result is the formal adoption of REACH-style chemical regimes in Canada, Australia, China, South Korea, Japan, and Russia; of GDPR-style data protection regimes in more than thirty jurisdictions; of RoHS-style hazardous substance restrictions in California, China, Japan, and South Korea. Each represents not a treaty-mediated convergence but a market-mediated one, codified after the fact.
The two effects are sequential. De facto adjustment paves the way for de jure adoption — not the reverse. This sequencing is one of the article’s most analytically important observations, because it inverts the implementation problem familiar to international legal scholarship. When a treaty requires firms to change their conduct, compliance is uncertain because the firms have not pre-committed; the gap between formal adoption and practical implementation is precisely where international agreements fail. In the Brussels Effect sequence, formal legal adoption merely codifies a market reality already established. The implementation problem is solved before the legislation passes. The law catches up with the practice; it does not have to create it.
The article is emphatic that the de jure effect is not necessary for the Brussels Effect to operate. The Effect is essentially about one jurisdiction’s capacity to override others — not about its capacity to trigger a formal race to the top. Bradford observes that the existing literature on the California Effect had tended to focus on the rarer and more visible de jure dimension — formal legislative adoption by other states of the lead state’s rules — and in doing so had underestimated the breadth of the underlying phenomenon. Global commerce, in many domains, already operates under unilateral EU rules even where other jurisdictions formally maintain their own laws. The U.S. retains its own privacy regime, its own competition rules, its own chemical regulation framework; yet in each of these areas, U.S. firms doing global business operate under European rules by economic necessity. The de facto dimension is where the action is. The de jure dimension is where, when it occurs, the practical reality is formally acknowledged.
This insight also illuminates why scholarship had previously thought the California Effect was constrained to “only a highly limited subset of environmental laws.”15 Earlier analyses had counted only the cases where foreign legislatures formally adopted Californian rules. They had not counted the cases where firms operated globally under Californian standards without any formal foreign adoption. The same blindness, Bradford argues, had obscured the much larger phenomenon of the Brussels Effect.
The case studies as proof of mechanism
Part II of the article supports the theoretical framework with five extended case studies. Four are clear instances of the Brussels Effect operating fully; the fifth — food safety and genetically modified organisms — is a partial case that illuminates the limits of the framework by showing what happens when the conditions are only partially satisfied.
The antitrust case is built around legal nondivisibility. Bradford’s principal example is the 2001 Commission decision to block the General Electric–Honeywell merger after the United States Department of Justice had already cleared it.16 The transaction was a forty-two-billion-dollar acquisition between two American companies, but because a merger cannot legally proceed in one jurisdiction and be enjoined in another, the strictest competition regulator’s veto became the global outcome. The earlier Boeing–McDonnell Douglas case followed the same logic: the Commission did not block the merger but extracted concessions, including the abandonment of exclusive dealing contracts with U.S. carriers, that the U.S. authorities had found unobjectionable. The pattern extends through the EU’s prohibition of the Aerospatiale–Alenia acquisition of de Havilland (approved by Canadian regulators), record-high fines against Microsoft and Intel, and the imposition of behavioural and structural remedies against dominant firms across multiple sectors and continents.17
Bradford observes that the asymmetry between the United States and the European Union in antitrust enforcement is not one of regulatory capacity. Both U.S. and EU agencies enjoy extraterritorial jurisdiction under the effects doctrine; both can apply their laws to foreign firms whose conduct produces anticompetitive effects within their territories. The asymmetry is one of regulatory preference. The Commission is suspicious of the market’s ability to deliver efficient outcomes and is therefore more inclined to intervene; the United States gives greater weight to the risk of inefficient intervention. The EU regulator emphasises the costs of false negatives — anticompetitive practices that go unaddressed — while the U.S. regulator emphasises the costs of false positives — procompetitive practices that are erroneously restricted. In a contest between the two preferences over the same transaction, legal nondivisibility transforms the more interventionist preference into the global outcome. The Commission has one additional structural advantage: it can prohibit mergers and impose behavioural and structural remedies without first obtaining court endorsement, whereas U.S. agencies must seek federal court approval to enjoin a merger.18
The privacy case turns on technical nondivisibility. Data flows are intrinsically transnational; the architecture of the internet does not permit a firm to maintain a European privacy policy distinct from its policy elsewhere when the same servers, the same code, and the same data structures serve both populations. Bradford traces this dynamic from the 1995 Data Protection Directive through the negotiation of the Safe Harbor Principles — the framework under which U.S. firms agreed to comply with EU privacy rules in exchange for the EU’s recognition that the resulting data flows met the “adequate level of protection” standard — and into the diffusion of EU-style privacy laws to more than thirty jurisdictions by the time of writing.19 By 2012, the EU was preparing the General Data Protection Regulation, which entered force in 2018 and has since become the canonical exemplar of the Brussels Effect, with GDPR-style frameworks adopted in jurisdictions as diverse as Brazil, India, Japan, South Korea, California, and South Africa.
The mechanism in the privacy case operates through both technical and economic nondivisibility. Technical: it is often impossible for a multinational firm to determine whether a particular user is an EU citizen, and the cheapest compliance strategy is to apply the European standard to all users. Economic: maintaining a European-only data architecture, where it is technically feasible, requires duplicate systems, duplicate compliance, and duplicate engineering effort. Bradford documents the resulting global adjustments by Google, Microsoft, General Motors, DuPont, Procter and Gamble, and others — firms that have a single global privacy policy, and it is Europe’s.20
The chemicals case centres on REACH — the Registration, Evaluation, Authorization, and Restriction of Chemicals regulation enacted in 2007. REACH places the burden of proof on manufacturers and importers, requiring them to gather and submit safety information on tens of thousands of chemical substances; it applies not only to new chemicals but also to “existing substances” already on the market; and it is guided by the precautionary principle, which lowers the threshold for regulatory intervention compared to the cost-benefit framework that governs the U.S. Toxic Substance Control Act.21 The U.S. TSCA grandfathers approximately ninety-five per cent of existing chemicals and places the burden of proof on regulators rather than manufacturers; the gap between the two regimes is enormous, both in stringency and in regulatory philosophy.
REACH propagates through economic nondivisibility. The chemical industry is structurally global; reformulating products to meet REACH standards and maintaining a parallel non-REACH-compliant production line for non-EU markets is rarely cost-rational, given the scale economies of unified production. The result is that REACH effectively governs a substantial fraction of chemical production worldwide. Bradford documents Dow, Ikea, Lego, Mattel, Revlon, Unilever, L’Oréal, and Estée Lauder among the firms that have applied REACH standards across their global operations.22 She also documents the secondary de jure cascade: California’s incorporation of REACH-derived standards into state law (the California Department of Toxic Substances Control is now required to use “to the maximum extent feasible” the safety information generated by REACH), state-level regulatory reforms in Maine and Massachusetts, congressional bills to amend the TSCA, and the adoption of REACH-style frameworks in Canada, Australia, China, South Korea, Japan, and Russia.
The environmental case illustrates economic nondivisibility in product design and operational compliance. The Restriction of Hazardous Substances Directive, enacted in 2003 and progressively extended, bans the release of hazardous substances into the environment when household appliances, computers, and other electrical and electronic products reach the end of their useful life. The Directive has prompted global redesign of electronic products, with multinational manufacturers preferring a single RoHS-compliant design for worldwide distribution rather than maintaining a parallel non-compliant line for non-EU markets. RoHS-style laws have been adopted in China, Japan, and South Korea, and California’s Electronic Waste Recycling Act of 2003 — the so-called “Cal RoHS” — explicitly incorporates EU standards and, strikingly, states that future amendments to the EU directive will be incorporated into California law.23
The most recent environmental example in Bradford’s account is the EU’s inclusion of foreign airlines in its emissions trading scheme, effective January 2012. All airlines operating flights into or out of EU airports must purchase emission permits for the full flight, not merely the portion in European airspace. On a flight from San Francisco to London, only nine per cent of emissions occur over EU territory, but the airline must acquire permits for the entire journey. The case generated immediate U.S. resistance — United, Continental, and American Airlines challenged the scheme before U.K. courts — but the European Court of Justice confirmed the Directive’s validity in December 2011, holding that the ETS does not violate principles of territoriality or third-state sovereignty because jurisdiction attaches at the point of landing within EU territory.24 The case extends the Brussels Effect into climate regulation, and the structure of the scheme — its “equivalent measures” exception — creates explicit incentives for foreign jurisdictions to adopt their own carbon pricing in order to be exempted from EU rules.
The food safety case is the partial one. The EU’s regulation of genetically modified organisms differs sharply from the U.S. approach: GMOs in the United States are treated as substantially equivalent to conventional products and can be cultivated and marketed without specific premarket safety studies or labelling, while the EU subjects GMOs to extensive premarket evaluation, applies the precautionary principle, and requires labelling of foods and feeds containing more than 0.9 per cent GMO content.25 Bradford treats the GMO case as a useful test of the framework because it satisfies some of the conditions for the Brussels Effect but not all of them clearly. The EU has the regulatory capacity and the preference for stringent regulation; the targets are inelastic consumer markets. But market power is contested — the EU is only the fifth-largest export market for U.S. farmers, accounting for roughly eight per cent of U.S. agricultural exports, which means many producers can plausibly divert trade elsewhere. And divisibility is contested in a different way: in principle, U.S. farmers could cultivate both GMO and non-GMO varieties for different markets, but in practice the segregation of seeds, growing areas, equipment, and supply chains is costly, and pollen drift between fields creates technical contamination risks. The result is partial nondivisibility, generating a partial Brussels Effect: many U.S. farmers cultivate only EU-approved GMO varieties because the costs of segregation exceed the benefits, but a substantial portion of U.S. agricultural production remains in non-EU-approved varieties.26 The case shows the framework at work in the negative — when the conditions are partially satisfied, the Effect is partially generated, and the structural argument of the article is reinforced by the imperfect fit of the imperfect case.
What drives the EU — external and internal motivations
Part III of the article addresses the question that the framework leaves open: why does the EU pursue regulatory globalisation in the first place? Bradford distinguishes between external motivations — what the EU seeks to achieve in the international system — and internal motivations — the domestic political and bureaucratic forces that produce the regulatory output that the international system then absorbs.
The external account begins by setting aside the most provocative framing. Some commentators have characterised the EU’s regulatory reach as a form of imperialism — a “regulatory imperialism” or “cultural imperialism” through which Europe imposes its preferences on the rest of the world using economic and bureaucratic tools where it lacks the military and political instruments of the older empires.27 Bradford finds this framing unpersuasive. The EU does not impose its rules on other states; it enforces its rules within its own market, applying them equally to domestic and foreign actors, and the global propagation that results follows from the rational choices of foreign firms rather than from any direct exercise of coercion. Whether the consequence is imperial in some normative sense is, in Bradford’s account, a less productive question than whether the mechanism is coercive in a structural sense — and structurally, it is not.
The EU’s own self-presentation locates its regulatory authority within a tradition of normative power. The European Commission’s 2007 policy paper A Single Market for Citizens envisioned the EU as a “standard-setter at the international level,” using its internal market as a launchpad for a “global agenda” of rules and standards in product safety, environmental protection, securities regulation, and corporate governance.28 The treaty language is in the same register: Article 3(5) of the Treaty on the Functioning of the European Union commits the EU to promoting its values in international relations, contributing to peace, sustainable development, the rule of law, and the protection of human rights. The presentation casts the Brussels Effect as the byproduct of a benevolent project — universal values, advanced through legitimate means, producing global public goods like climate mitigation and consumer protection.
Bradford is sympathetic to this framing without fully accepting it. She acknowledges that the EU genuinely sees itself as advancing universal norms, and that the climate regulation case in particular has a clear public-goods rationale: climate change is a global problem, the EU has limited capacity to mitigate it acting alone, and unilateral regulation is one of the few instruments available given the difficulty of negotiating effective international treaties. But she also notes that any actor seeking to shape the international order must use the means available to it, and regulatory power is what the EU has. The distinction between benevolent norm projection and self-interested unilateralism is in practice often impossible to draw cleanly, and the analytical move that matters is not to adjudicate the EU’s motives but to specify the mechanism through which whatever motives produce regulation produce, in turn, global regulatory consequences.
The internal account is, in Bradford’s framework, the more powerful one. The EU’s regulatory authority was built primarily for internal purposes — to create and protect a single market — and its external reach is to a large extent a byproduct of that internal project. Three internal dynamics are particularly important.
First, the EU has a powerful interest in protecting the competitiveness of its own firms. European firms operate under demanding regulatory regimes; allowing foreign firms to access the European market under more lenient foreign standards would put European firms at a competitive disadvantage. By requiring foreign firms to meet European standards as a condition of market access, the EU levels the playing field at home. By generating the Brussels Effect that propagates those standards globally, it levels the playing field abroad. The inclusion of foreign airlines in the emissions trading scheme was justified in precisely these terms — the Commission spokesman explained that the EU “can’t impose a burden only to European airlines and not include others” without distorting competition.29
Second, the dynamic operates within the EU itself before it operates globally. Strict member states — Germany, the Netherlands, the Nordic countries in environmental regulation; France and Germany in privacy — lobby for EU-level standards that match their domestic preferences, levelling the playing field against firms based in less stringent member states. Once the standard is harmonised at EU level, it becomes the platform from which the Brussels Effect projects outward. The Effect is therefore double-layered: it operates first within the Union to consolidate the strict standard, and then beyond it to globalise the consolidated standard.
Third, an unusual political coalition supports the externalisation of EU rules. Consumer and environmental advocates benefit from broader adherence to standards they consider normatively desirable. European firms benefit from the imposition of those same standards on foreign competitors. The two constituencies, often opposed on domestic regulatory questions, frequently find themselves aligned on the question of whether to push EU standards onto the rest of the world. Bradford gives the example of the EU’s Eco-Management and Auditing Scheme, which united European corporations and environmental NGOs in a successful campaign to convert EU disclosure requirements into International Organization for Standardization standards adopted globally.30
A final internal driver Bradford identifies is bureaucratic. The European Commission has expansive legal competence over the single market and limited competence over foreign and security policy, where decisions typically require unanimity among twenty-seven member states. Regulation is therefore the area where the Commission’s autonomous capacity for international action is greatest. Expanding the regulatory frontier is, among other things, a bureaucratic strategy for expanding the Commission’s role in world affairs through the instruments over which it actually controls. The external regulatory power of the EU is, in this account, partly a function of the structural distribution of competences within the Union itself.
Unilateralism, multilateralism, and the choice between them
Part IV of the article asks why the EU sometimes prefers unilateral regulatory action and sometimes engages in multilateral negotiation. The answer determines a great deal about where the Brussels Effect operates and where it does not.
Unilateralism has several relative advantages for a regulator whose preferences exceed those of its negotiating partners. It is faster — international negotiation can take decades, and outcomes are typically the lowest common denominator that all parties can accept. It is more stringent — the EU does not have to dilute its preferred rules in order to secure agreement from less ambitious states. It preserves regulatory autonomy — the EU is not bound by reciprocal commitments that might constrain its future flexibility. And, under the conditions Bradford identifies, it is more effective at actually producing global regulatory change: where the five conditions hold, unilateral action produces de facto convergence, while multilateral negotiation often produces formal agreement without practical implementation.
These advantages explain why, in areas where the conditions for the Brussels Effect hold most strongly, the EU has tended to act unilaterally rather than pursue multilateral solutions. The EU has not negotiated international treaties to harmonise privacy standards; it has enacted the GDPR and let the Brussels Effect do the diffusion. It has not pursued international agreements to harmonise chemical regulation; it has enacted REACH. It has not negotiated multilateral antitrust convergence through the World Trade Organization or the OECD; it has continued enforcing its own competition rules with extraterritorial reach.
Multilateralism persists, however, for several reasons. Some domains are not amenable to unilateral action — climate change requires global coordination, certain trade and security matters cannot be addressed through market access alone — and the EU pursues international agreements in those areas as a necessary complement to its unilateral regulatory power. Some forms of legitimacy require multilateral consent — the EU’s claim to be advancing universal norms is more credible when those norms are formally endorsed by other states through international agreement, and the EU therefore has reputational incentives to engage in multilateral processes even where unilateral action would suffice substantively. And the EU’s commitment to a rules-based international order is a constitutive feature of its self-understanding; multilateral institutions are part of the order it seeks to maintain, regardless of their instrumental utility on any particular question.
The result is a hybrid strategy: unilateral regulation where the conditions for the Brussels Effect hold, multilateral negotiation where they do not, and complementary use of both instruments in many domains. The EU does not face a choice between unilateral and multilateral regulatory globalisation; it deploys both, and the analytical question is when one is preferred to the other.
The limits — external impotence, internal fragility
Part V of the article is, for the researcher working in regulatory interoperability, perhaps more important than the affirmative theory. Bradford devotes substantial attention to mapping the constraints on the Brussels Effect, and the structure of those constraints determines where the Effect can and cannot operate.
External constraints are weak. This is the somewhat unsettling conclusion that emerges from a careful reading of Part V.A. Three potential checks on the EU’s regulatory power — markets, other states, and international institutions — turn out to be largely impotent against it.
Markets cannot discipline the Brussels Effect because the EU primarily regulates inelastic targets. Consumers cannot relocate to lax jurisdictions in protest of strict standards. Firms can, in principle, divert exports away from the European market, but for products with affluent and inelastic European demand, the opportunity cost of doing so is prohibitive. The market signal that would, in some other context, punish excessively stringent regulation by triggering capital flight or trade diversion is, in the Brussels Effect cases, structurally muted.
Other states cannot effectively counter the Effect either, and this is perhaps the most counterintuitive of Bradford’s claims. The intuitive response of a state whose preferences are overridden by EU rules would be to escalate — to adopt still stricter regulations of its own. But this is precisely the wrong move. Outpacing the EU through still stricter regulation only leaves the counter-regulating state with even less desirable standards than it would otherwise have. The structure of the regulatory race ensures that the stricter regulator wins, but it also ensures that no rational state wants to win such a race when it would not have chosen to be that strict in the first place. The result is a kind of regulatory paralysis: states whose preferences are overridden have no good response except to accept the EU’s standards, or to attempt to negotiate (which the EU can decline) or to challenge through international institutions (which, as below, are largely ineffective).
International institutions are similarly impotent. The World Trade Organization’s jurisdiction does not extend to most of the areas where the Brussels Effect operates. Privacy regulation, antitrust, broad swathes of consumer protection — these fall outside the WTO’s core competences. Even where the WTO does have jurisdiction, as in some aspects of food safety and technical barriers to trade, its dispute settlement system has shown itself unable or unwilling to discipline EU regulatory choices that have global effects. The 2006 GMO dispute, in which the WTO ruled against the EU’s de facto moratorium on biotech approvals, has produced no practical change in EU policy and no resumption of significant transatlantic GMO trade.31 The lesson Bradford draws is that international institutions, designed to address trade barriers and discriminatory policies, are poorly equipped to discipline the global externalities of internally-coherent regulatory choices.
Internal constraints, by contrast, are real and important. Bradford identifies two principal ones.
The first is variation in EU regulatory competence. The Brussels Effect can only operate in domains where the European Union, by treaty, has the legal authority to produce binding rules. In domains where member states have retained competence — foreign and security policy, direct taxation, significant areas of social policy and labour regulation — the EU as such cannot generate Brussels Effects because it cannot produce binding rules. This is why the Effect is concentrated in single-market matters, environmental policy, competition policy, and consumer protection, and why it does not extend to, say, defence procurement or income taxation.
The second internal constraint is the political economy of the EU itself. As the Effect expands and as more domains come under EU regulatory competence, the costs of stringent common standards fall unevenly on different member states. Member states with more permissive domestic preferences bear adjustment costs that those with stricter preferences do not. Member states with export industries that benefit from common standards see different incentives than those whose firms are net importers. Member states with more sceptical publics face political costs in implementing EU rules that more enthusiastic publics do not impose on their governments. The expansion of competence and the diffusion of effect are, in the long run, in tension with the political consent that sustains the regulatory machinery in the first place. Bradford’s expectation, as of 2012, was that this internal tension would be the most important brake on the Brussels Effect over time. The years since have confirmed that prediction more vividly than the article itself anticipated, with the politics of EU regulatory expansion — including the politics of the AI Act — turning on precisely these internal divisions.
There are also areas, Bradford acknowledges, where the structural conditions do not hold and where the Brussels Effect therefore cannot operate regardless of EU political preferences. Labour standards are divisible across markets and so escape the Effect. Financial transactions are elastic and so cannot be effectively regulated by a single jurisdiction without diversion to other financial centres. Tax competition between jurisdictions remains a structurally Delaware-Effect domain rather than a California-Effect domain, with capital mobility ensuring that downward harmonisation pressures exceed upward ones. These structural edges of the framework are as important to the analytical theory as the cases where the Effect operates fully.
The framework's afterlife — and the AI Act question
Bradford does not discuss artificial intelligence — the AI Act would not be tabled for another nine years — but the framework she developed in 2012 is now the dominant lens through which scholars and policy-makers analyse the global reach of EU AI regulation. For doctoral research on regulatory interoperability between the AI Act and the NIST AI Risk Management Framework, the question is whether the five conditions hold in the AI domain. The answer is more complicated than the framework’s celebrated success in GDPR would suggest.
The AI Act, in force from 2024, was explicitly designed in the shadow of the Brussels Effect. Its drafters anticipated, and the secondary literature has largely concluded, that the Act will project EU standards onto AI systems deployed globally through the same mechanisms Bradford identified for GDPR.32 A frontier model trained for deployment in the EU must satisfy EU obligations on documentation, transparency, risk assessment, human oversight, conformity assessment, and post-market monitoring; the architectural and economic costs of maintaining a parallel non-EU-compliant version of the same model are, for most providers, prohibitive. The de facto cascade should follow the GDPR pattern.
Whether it will, however, depends on the precise satisfaction of all five conditions, and the AI case strains several of them in ways that the GDPR case did not.
Market power in the AI domain is more contested than in the privacy domain. The EU remains a major market for AI services, but its share of global AI demand is structurally smaller than its share of, say, chemical demand or consumer goods demand. The leading frontier developers are concentrated in the United States and, increasingly, in China; the EU itself has produced relatively few frontier-scale labs (Mistral being the principal European entrant). For an American or Chinese frontier developer weighing the costs of EU compliance against the opportunity costs of EU exclusion, the calculation is genuinely closer than it would be for a cosmetics manufacturer or a chemical producer. Several major U.S. firms have already announced delayed or reduced EU deployments in response to AI Act obligations, suggesting that market power is, in the AI case, not as overwhelming as it has been in other Brussels Effect domains.
Regulatory capacity is novel and largely untested. The AI Act’s enforcement architecture is distributed across the European Commission’s new AI Office, national competent authorities in each member state, and a coordinating European Artificial Intelligence Board. None of these institutions has the accumulated regulatory expertise that the Commission’s DG Competition has developed over decades of antitrust enforcement, or that the European Chemicals Agency has developed for REACH. The capacity exists on paper; whether it will translate into the kind of credible enforcement that produces firm adjustment is a question that will only be answered over the next several years.
Preference for strict rules is present but internally contested. The AI Act passed with significant disagreement among member states about its stringency. France, Germany, and Italy resisted aspects of the foundation-model regulation during the trilogue, fearing competitive disadvantage for European labs. The political economy of implementation is genuinely fragile, with periodic calls from European industry and some member-state governments for delays, exemptions, or relaxation. The third condition holds, but more weakly than it has historically held in privacy or chemical regulation.
Inelastic targets hold in mixed form. The end-users of AI services are largely inelastic — consumers and businesses operating in the EU do not relocate to escape the AI Act. But the developers of AI services are partly elastic in a way that consumer goods manufacturers are not. A frontier model can be withheld from the EU market — not deployed, not made available — at relatively low cost to the developer, who continues serving the rest of the world. Several U.S. firms have done exactly this with specific products, choosing exclusion over compliance. The structural anchor that ensured the Brussels Effect in other domains — the impossibility of forgoing the European consumer — is, for AI services, partial.
Nondivisibility holds for some systems and not for others. Foundation models trained on a single architecture are economically and technically difficult to bifurcate; the cost of maintaining a separate EU-compliant model would be substantial. But jurisdictionally segmented deployments are technically feasible for many narrower applications, and the architecture of the AI value chain — with foundation models, fine-tuned models, deployed applications, and end-user interfaces forming a layered stack — allows for compliance to be concentrated at one layer (typically the deployment layer) without propagating to others. The result is a partial nondivisibility that may, like the GMO case, generate a partial Brussels Effect rather than a complete one.
The literature on regulatory interoperability between the EU AI Act and the NIST AI Risk Management Framework — the principal U.S. soft-law instrument addressing comparable risks — sits squarely on this analytical terrain. If the Brussels Effect operates fully in the AI domain, interoperability becomes a question of how American institutions accommodate the European standard, since the European standard will, by economic necessity, govern most globally deployed systems. If it operates only partially, interoperability becomes a more genuinely bilateral problem: mapping two distinct regulatory architectures onto a shared technological substrate in a way that allows U.S. firms operating under NIST RMF to demonstrate substantial compliance with AI Act requirements without duplicating control structures, and vice versa.
The doctrinal stakes are not abstract. If a U.S. federal agency operating under the NIST AI RMF cannot, by virtue of classification restrictions and the soft-law character of its compliance regime, demonstrate substantive equivalence to AI Act conformity assessment, then either the Brussels Effect drags American regulatory practice toward EU forms (the strong Brussels Effect scenario), or the two regimes diverge with friction (the weak Brussels Effect scenario), or interoperability protocols are negotiated that allow recognised equivalence without full alignment (the negotiated scenario). Each of these scenarios maps to a different reading of how the five Bradford conditions hold in the AI domain. The framework remains the analytical instrument with which the question must be asked, but the answer it produces is no longer the GDPR-style strong Brussels Effect that the framework’s first generation of applications suggested. The answer is more contingent, more contested, and more dependent on the political economy of the EU’s own implementation than on the structural force of its market power alone.
Coda — the framework, fifteen years on
What remains theoretically robust in Bradford’s argument is the five-condition framework itself. Fifteen years of subsequent scholarship has tested it across new domains. In financial regulation, where targets are elastic and nondivisibility weak, the Effect does not operate — confirming the framework’s negative predictions. In data protection, where technical and economic nondivisibility are total, the Effect operates almost completely — confirming the strong positive predictions. In content moderation, where divisibility is partial and political contestation high, the Effect operates with significant friction. In artificial intelligence, where the answer is still being written, the framework predicts a partial Effect — and the early evidence is consistent with that prediction. The framework has held up across this empirical expansion, and the analytical move that gave the article its lasting importance — the disaggregation of de facto and de jure mechanisms, and the identification of nondivisibility as the operative conversion factor — has become canonical in the international regulatory literature.
What has aged, less well than the structural argument, is the article’s implicit normative posture. Bradford in 2012 wrote in a register that was, if not celebratory, at least curious and admiring — describing a form of European power that had been overlooked, was real, and was producing measurable global effects in domains where international institutions had failed. The decade and a half since has complicated that posture. The Brussels Effect is now visible enough to its targets to provoke counter-mobilisation. In the United States, the response has taken the form of state-level data-protection laws diverging from GDPR templates rather than adopting them, federal legislative resistance to importing EU-style rules, and the assertion of an alternative American regulatory model in Digital Empires terms. In China, the response has been the construction of an explicit Chinese alternative to both American and European digital governance models, with its own technical standards, its own data localisation regime, and its own forms of regulatory projection through Belt and Road infrastructure and digital platform exports. In major developing economies — India, Brazil, Indonesia — the response has been selective adoption of EU standards combined with active resistance in domains where the political cost is higher than the regulatory benefit.
The framework Bradford built in 2012 did not contemplate these counter-strategies because, in 2012, they were not yet visible. The Effect was at its empirical peak, the EU’s regulatory output was at its most ambitious, and the targets of regulation had not yet learned to anticipate the mechanism. Fifteen years later, the targets have learned. They have built institutions designed to resist. They have negotiated equivalence and adequacy regimes that allow them to maintain regulatory autonomy while accessing the European market. They have, in some cases, simply chosen exclusion over compliance, accepting the loss of EU access as the price of their own regulatory preferences.
The reader is left with a question that Bradford did not need to ask in 2012 and which the architecture of her framework does not directly answer: what happens to unilateral regulatory globalisation when its targets recognise the mechanism, learn to anticipate it, and begin to build institutions designed to resist it. The five conditions she identified remain individually necessary and jointly sufficient as a matter of structural logic. But the political conditions for those conditions to keep holding may be more fragile than they were when the article first appeared. For the doctoral researcher working on EU-US regulatory interoperability in AI, this is the analytical horizon. The Brussels Effect remains the frame. But the frame itself is now under pressure, and the question of whether a U.S. federal agency operating under the NIST AI RMF can demonstrate substantial compliance with the EU AI Act is, in significant part, a question about whether the Effect that gave the framework its empirical purchase will continue to operate in a domain that was unimagined when the framework was written.
What endures from Bradford’s article, beyond the conditions and the cases, is a methodological lesson. The dominant accounts of power in international relations had focused on military force, economic sanction, and the conditional incentives of international institutions; Bradford showed that an entire register of power had been operating in plain sight, was largely invisible to that literature, and could be specified with analytical precision once the right framework was applied. The framework she built is now the standard tool. The lesson that the standard tools of a discipline can fail to capture the most important phenomenon in front of them is the deeper lesson, and it remains pertinent whenever a new technology, a new domain of regulation, or a new political configuration emerges that the existing categories were not designed to describe.