Sovereign AI

Sovereign AI is a country’s or organisation’s capacity to build and run AI on its own terms: local compute, data, models and talent, under its own law, rather than depending on foreign providers. Motivations include national security and control of public-sector systems, and local language and culture (per Wikipedia); data residency, privacy and resilience against supply or policy shocks are also commonly given but not sourced here. Approaches range from state-funded national models and data centres, to regional clouds and “sovereign cloud” offerings from hyperscalers, to enterprises self-hosting open-weight models.

Links

• Compute: sovereign programmes drive demand for GPU capacity; see neocloud and nvidia.
• Models: open-weight families (mistral-family, deepseek-family, gemma-4) are used as a basis; mistral-ai is described in the vault as a European option.
• Law and privacy: ai-regulation-and-policy, data-privacy-in-ai.
• Local running: ollama, vllm, inference-and-serving.

Related: foundation-and-frontier-models, scaling-laws, ai-research-labs.

Sources

• https://en.wikipedia.org/wiki/Sovereign_AI (secondary; re-fetched 2026-10-07: loosely defined policy term; motivations national security, control of public-sector systems, local language and culture; approaches include local data centres, language-specific models, domestic investment, procurement; sovereignty need not mean full domestic production)

Open items

• Motivation ‘resilience against supply or policy shocks’ and ‘data residency and privacy’ are common usage but not stated in the Wikipedia summary; unsourced.
• Country programmes and spending figures are intentionally omitted; add with dated primary sources.