Azure AI Foundry
External APIs
Azure AI Foundry, Microsoft's unified platform, provides SaaS access to a broad catalogue of models, including OpenAI's GPT, Llama, Mistral and Phi, behind a managed API. European regions exist, including France Central and West Europe, which brings data hosting closer to home. The vendor remains a US entity, and is therefore subject to the Cloud Act. The platform uses it for non-sensitive innovation.
Key points: OpenAI and open models · Managed · EU regions · Non-sensitive.
Models served: GPT-5.6 and GPT-5.5 from OpenAI (GPT-5.4, GPT-4.1 and o4-mini remain in the catalogue), Claude from Anthropic, Llama from Meta, Mistral, Phi from Microsoft, DeepSeek V4 and Grok 4.6 from xAI (Microsoft catalogue as of 2 September 2026)
Location: United States, EU regions available, subject to the Cloud Act.
Where the data goes
This is the point of the offering. A European Union Data Zone deployment processes prompts and responses only within Microsoft's EU Data Boundary, which can include EFTA countries such as Norway and Switzerland in addition to EU member states; data at rest stays in the chosen Azure geography (Microsoft documentation as of 2 September 2026). GPT-5.6 and GPT-5.5 can be deployed there in a data zone from France Central, Sweden Central or West Europe (regional availability as of 2 September 2026).
What residency does not solve
Data processed in France by a provider under US law remains reachable under the Cloud Act: residency moves where processing happens, not the jurisdiction of the parent company. It is a real risk reduction on latency, on GDPR and on transfers. It is not sovereignty.
When to use it from a sovereign platform
When you need OpenAI models with contractual European processing, and Cloud Act exposure is accepted knowingly. For processing that must escape that exposure, only the local foundation answers.