Google Cloud Sovereign AI in Europe
5 étapes1 pièces4 Sources
Google Cloud is attempting to rebuild trust with European public sector and regulated enterprise customers by introducing technical isolation measures for its sovereign AI offerings. At the Sovereign AI Summit in Brussels on November 17, 2026, Google announced that its flagship Gemini models will now be available on Google Cloud Dedicated, a partner-operated cloud infrastructure designed to remain operational for up to 12 months even if disconnected from Google's central systems. This move addresses growing European anxiety over political risks and potential US government interference, particularly under the Trump administration's "America First" policies.
The core of Google's strategy relies on three architectural tiers: Google Cloud Data Boundary (public cloud with strict residency controls), Google Cloud Dedicated (isolated infrastructure run by local partners like S3NS in France and Thales in Germany), and Google Distributed Cloud (on-premises or air-gapped). Key announcements include the preview of Confidential External Key Management (cEKM), which allows customers to hold encryption keys outside Google's control, and new audit frameworks mapping to EU regulations like DORA. However, skepticism remains. German Chancellor Friedrich Merz recently endorsed Google Cloud for intelligence services, calling it "safe," but critics argue this ignores the "soft kill-switch" risk where software updates could be halted by US authorities. Meanwhile, the pending EU Cloud and AI Development Act (CADA) may still restrict market access for non-EU providers regardless of technical mitigations.
Comment on en est arrivé là
May 2025: Google introduces its initial portfolio of sovereign cloud offerings.
cloud.google.com ↗February 2026: Google sets out five digital sovereignty commitments, including local investment and operational continuity.