Affiliation: Professor of Law and International Organization, Columbia Law School; author of The Brussels Effect (2020) and Digital Empires (2023) One-line position: The EU's regulatory model will shape global AI rules whether other nations like it or not — the "Brussels Effect" means companies adopt EU standards worldwide to avoid fragmentation, making the AI Act a de facto global regime.
What she's reacting against
- US-centric framing of AI governance as a US-China bilateral competition — argues the EU is the regulatory third pole
- The assumption that regulation kills innovation — argues the EU model creates predictable rules that enable investment
- Tech industry narratives that self-regulation is sufficient — insists democratic governance of AI requires statutory frameworks
- The view that the EU is merely reactive — argues it actively shapes global market standards through its regulatory ambition
Key claims
- The Brussels Effect: the EU's ability to project its market standards globally; companies adopt EU rules for their entire operations (not just EU sales) because maintaining dual standards is too costly
- The AI Act will have Brussels Effect dynamics: risk-based classification of AI systems will become the de facto global standard, even in countries without equivalent legislation
- Three competing "digital empires" (Digital Empires, 2023): the US (market-driven), EU (rights-driven), and China (state-driven) models are competing visions for governing technology, including AI
- Regulatory export is structural, not political: it happens through market mechanisms (companies prefer one standard), not diplomatic pressure
- Democratic governance of AI is non-negotiable: AI systems that affect fundamental rights require democratic oversight — this is a values claim, not just an efficiency claim
Theories aligned with
- Regulatory globalization / Brussels Effect theory
- Rights-based technology governance
- Adjacent to Khan's antitrust frame but operating at the international regulatory level rather than domestic enforcement
Where she overlaps / splits
- Overlaps with Schaake on EU governance leadership; splits on register — Bradford is the academic theorist, Schaake is the practitioner/policy operator
- Overlaps with Khan on AI needing structural regulation; splits on mechanism — Khan uses antitrust enforcement, Bradford uses regulatory export
- Overlaps with Mazzucato on state having a legitimate role in shaping technology markets; splits on focus — Mazzucato is industrial policy, Bradford is regulatory standard-setting
- Splits with Andreessen/Thiel on regulation's effect on innovation — they see it as drag, she sees it as enabling structure
- Splits with US tech policy (broadly) on the adequacy of self-regulation
Notable predictions
- (2020) The Brussels Effect would extend to digital regulation — outcome: supported (GDPR adopted globally, AI Act influence growing)
- (2023) Three competing regulatory models (US/EU/China) would define AI governance — outcome: supported by divergent approaches to AI regulation
- (Ongoing) The AI Act will be adopted beyond Europe through the Brussels Effect — outcome: TBD; early evidence in India and other jurisdictions studying the EU model
Track record
- The Brussels Effect (2020) coined and validated a concept now used across policy discourse — genuine field-defining contribution
- Digital Empires (2023) named FT best book of 2023; won 2024 Stein Rokkan Prize — peer recognition is strong
- GDPR as prior evidence: the Brussels Effect thesis was validated by GDPR's global adoption before the AI Act was even drafted
- Advisory roles to EU institutions — influence on actual policy, not just academic commentary
Empirical vs normative
- Empirical: how regulatory standards spread across markets; case studies of GDPR, chemicals regulation, data protection as Brussels Effect precedents
- Normative: democratic governance of AI is a fundamental requirement; the EU model is preferable to market-driven or state-driven alternatives
- The empirical machinery (Brussels Effect as mechanism) is well-documented; the normative preference for the EU model is a values choice
Sources
- Books: The Brussels Effect: How the European Union Rules the World (2020); Digital Empires: The Global Battle to Regulate Technology (2023-09)
- Columbia Law School: faculty page with CV
- Project Syndicate: regular columnist
- GovAI: "The Brussels Effect and Artificial Intelligence" (research paper)
- Capgemini: interview on tech sovereignty
Weak spots / open questions
- The Brussels Effect assumes companies prefer one global standard — but AI regulation is complex enough that dual-compliance or market-specific models may emerge
- EU AI Act enforcement is untested — the regulatory ambition may exceed implementation capacity
- The EU model's preference for precaution over speed may cede frontier AI development to US and China — Bradford's frame doesn't fully address the innovation-cost tradeoff
- "Rights-driven" framing assumes consensus on which rights matter and how to weight them — this is contested even within Europe
- Less engaged with technical AI questions — the frame is regulatory structure, not model capability or alignment
Rich's take
- (your synthesis here)
converts-from: personas/anu-bradford.md · schema v1 · AI & Society domain