Weighted Score:
77
High sovereignty because French-headquartered AI developer under EU jurisdiction; however relies on global cloud and semiconductor supply chains.

Company name: Mistral AI SAS
URL: https://www.mistral.ai
eurotechguide review(s): Not reviewed

Digital services offered: Chatbot (GPT)
Criterion Score Short Assessment Long Rationale
SOV-1 (strategic) 4 EU HQ (France); EU legal control. Mistral AI is headquartered in Paris, France and governed under EU law. Strategic control resides in the EU, aligning with “subject only to EU law” under the framework’s legal axis. –>SEAL-4
SOV-2 (legal) 4 EU governance; no US parent. Corporate governance and ownership are EU-anchored. No inherent foreign extraterritorial regime exposure via ownership structure is publicly evidenced. –>SEAL-4
SOV-3 (data & AI) 4 Cloud/infra dependent on global providers. Central instance runs on Mistral Compute (a European-hosted AI cloud). Admins can enable audit logs; users can opt out of training use. Fine-tuning data is kept until you delete it. → SEAL-4
SOV-4 (operational) 3 EU deployable; infra choice determines control. Operational sovereignty depends on deployment environment. EU hosting is possible, but SaaS offerings may rely on non-EU cloud layers. –>SEAL-3
SOV-5 (supply chain) 2 Global semiconductor dependency. Advanced AI model training depends on globally sourced GPUs (e.g., US-designed semiconductors) and supply chains outside EU control. –>SEAL-2
SOV-6 (technological) 3 Partially open; ecosystem dependencies remain. Models are partially open-weight, improving transparency, but core proprietary systems and reliance on global ecosystems limit full technological sovereignty. –>SEAL-3
SOV-7 (security) 3 Structured security; limited certification transparency. Security posture appears structured, but no publicly evidenced EU sovereign-grade certification framework for services. –>SEAL-3
SOV-8 (environmental) 2 Limited verified AI sustainability reporting. Public sustainability metrics specific to AI infrastructure (PUE, water, renewable sourcing) are limited. –>SEAL-2

The scoring is done according to the Cloud Sovereignty Framework (v1.2.1 , October 2025). The scores are determined based on publicly available information, but something may have been missed. In case you want to challenge the score or submit a new digital service, click here