The Civic Node

AI Sovereignty

3 min read

Definition

AI sovereignty refers to nations’ efforts to establish control over, or domestic capacity in, AI development, infrastructure, and talent. It encompasses both offensive moves (recruiting foreign AI companies, building domestic AI capacity) and defensive moves (restricting foreign AI access, blacklisting non-compliant companies).

Why It Matters for the Newsletter

Power: AI is becoming a geopolitical battleground where nations compete not just militarily or economically, but for the infrastructure of intelligence itself. The Anthropic case shows both sides: the US trying to control AI deployment for military purposes, and the UK exploiting the resulting rift to attract AI talent.

DePIN: AI sovereignty pressures will hit AI infrastructure networks — including DePIN AI compute networks (e.g., Render, Akash) — in the same way they’re hitting Anthropic. The question of “who controls the compute” is a sovereignty question.

Evidence & Examples

  • UK recruiting Anthropic (2026): Britain offering London expansion + dual listing to capitalize on US-Anthropic conflict Britain woos Anthropic expansion after US defence clash
  • US blacklisting Anthropic: Using supply-chain risk designation to prevent a non-compliant AI company from operating with federal agencies
  • EU AI Act: Regulatory sovereignty — Europe setting rules for AI systems operating in its market
  • China’s AI regulation: Requiring LLMs to align with “core socialist values” — content-level sovereignty
  • On-device inference (2025–2026): On-Device AI extends sovereignty to the individual and team level. A 512GB Apple Silicon Mac can run DeepSeek R1 (671B parameters) locally. No API, no cloud provider, no data exposure. See Best Local LLMs for Every Apple Silicon Mac — 2026 Guide.
  • Trump federal AI EO postponement (May 21, 2026): A planned executive order with provisions covering Pentagon system security, federal civilian system security, federal AI tool use, and a voluntary pre-public-access framework with AI developers was postponed the day of signing. Trump’s stated reason: “I didn’t like certain aspects of it” — would “get in the way” of “leading China, leading everybody.” The pre-public-access provision is the candidate red-flag clause (OpenAI IPO window, Anthropic-Pentagon refusal track). No reissue timeline announced. Trump Postpones AI Executive Order — CBS - 2026-05-21
  • Newsom California state-level countermove (May 21, 2026): Same-day signing of California AI Workforce Disruption EO; 180-day deliverable on WARN Act updates for AI-driven layoffs. The federal-state sequencing means any eventual federal preemption language must be drafted against the now-published California text — Newsom’s order sits at the state-level high-water mark. The same week documents both ends of the federalism spectrum: federal EO postponed; state EO signed and published. Newsom AI Workforce Disruption Executive Order — CBS Sacramento - 2026-05-21

Tensions & Counterarguments

  • AI sovereignty may be impossible to fully achieve — AI development is globally distributed and hard to contain
  • Recruiting individual companies doesn’t create national AI capacity — you need talent, compute, and data ecosystems
  • Sovereignty measures can fragment AI development, slowing progress for everyone

Key Sources

Graph View

Backlinks