Digital Empires — The Tripartite Governance Frame

Idea

Per Anu Bradford (bradford-digital-empires), three distinct digital-governance models compete globally: the US market-driven model, the Chinese state-driven model, and the EU rights-driven model. Each exports a different form of power and produces a structurally different AI value stack (divergent-value-stack-optima).

The three models

ModelLogicExports
US — market-drivenFree markets, free speech, minimal pre-emptive regulation. Section-230-style liability shields enabled hyperscaler scale.Private corporate power — the platforms themselves.
China — state-drivenTechnology as instrument of state security, industrial policy, and global infrastructure projection.Infrastructure power — 5G, undersea cables, data centers, surveillance kit (Digital Silk Road).
EU — rights-drivenFundamental rights, democratic structures, fair redistribution of digital gains. Check corporate power.Regulatory power — the Brussels Effect (brussels-effect-and-mirage).

Why this matters for value-capture

Each model’s export mechanism dictates where value lands:

  • The US captures rent at the API meter because its model exports the platforms.
  • China captures rent at the system-integration layer in client states because its model exports physical and digital infrastructure into them.
  • The EU attempts to capture rent at the compliance gate because its model exports norms, not products.

This is the structural reason the AI value stack has no single shape — see divergent-value-stack-optima.

Bradford’s diagnosis of Europe

Bradford explicitly argues that Europe’s technological lag is not caused by its regulations. The deeper structural deficits are:

  • No truly integrated digital single market (27 markets, 24 languages).
  • No capital markets union; venture capital ecosystem orders of magnitude smaller than the US.
  • Cultural attitudes toward risk and failure that suppress founder formation.

Regulation is then a symptom, not a cause. This nuance pushes back against a simple “AI Act killed European AI” reading.

Tensions

  • The three-bloc frame treats geopolitical alignment as the master variable. It is silent on cross-bloc actors (India, Gulf, Southeast Asia, Latin America) who may pick and mix.
  • The “exports power” frame implies stable equilibria. The current era is one of active re-bordering — semiconductor export controls, data-localization rules, sovereign-AI initiatives — which could destabilize all three exports simultaneously.

Complications from OECD G7 evidence

The OECD G7 evidence (oecd-sme-ai-adoption-2025) partially confirms the tripartite frame at the level of rhetoric and regulatory architecture (US market-led, EU rights-led patterns clearly visible in the country profiles), but complicates it at the level of operational instruments and at the level of Japan/UK (see g7-sme-ai-policy-pluralism):

  • Japan is a fourth model, not a member of the existing three. State-coordinated + soft-law + normative leadership via Hiroshima AI Process — not market-led, not rights-led, not China-style state-controlled.
  • UK has diverged post-Brexit toward a Japan-like state-led + market-enabled posture (Modern Industrial Strategy, AI Opportunities Action Plan).
  • Canada bridges federal coordination with regional delivery and explicit compute-sovereignty framing.

A subtle but striking corroboration: in the OECD SME barrier survey, German SMEs report client disapproval as the top barrier to generative AI — a trust- and reputation-driven constraint, not a skills or money one. This is uniquely consistent with the rights-led EU posture and gives the tripartite frame an empirical SME-level anchor it didn’t have before.

Confidence: medium-low (downgraded from medium). At the level of regulatory architecture, the frame holds. At the level of operational policy instruments, G7 governments converge across the supposed empires. The frame may be more useful for legal/regulatory analysis than for predicting where capital and capability actually flow.