Sovereign AI in banking and finance
Sovereign AI is artificial intelligence whose data, models and operation stay under the control of the organization or country that uses it. Red Hat splits the idea into data sovereignty, technical sovereignty over the software stack, operational sovereignty over who runs and administers the system, and assurance that all of it can be verified.
For a bank the test is practical. If its AI assistant reads customer files, the bank has to know which law governs the model provider and whether it could keep running the model if the provider changed its terms. A hosted API from a provider under US law answers both questions differently from an open-weight model on servers in Germany. The page on AI infrastructure in banking compares the hosting options.
The Sparkassen run their own AI platform
Finanz Informatik, the IT provider of the German Sparkassen, runs S-KIPilot, an AI assistant platform for the savings banks. According to a report in IT Finanzmagazin (in German), the platform runs only inside Finanz Informatik's protected infrastructure, and data exchange with external systems is excluded by design. It uses open-weight models: Mistral 8x22B and Mistral Small, OpenAI's gpt-oss-120b for complex generation, Whisper for transcription and dots.ocr for reading documents.
The report names the choice of Nvidia hardware as the decision that made this possible, because a standard hardware base lets the team swap models as better ones appear. The Sparkassen-Finanzgruppe plans to train about 200,000 employees in AI skills.
BNP Paribas and Mistral AI
BNP Paribas took the route of a European model provider. In July 2024 it signed a multi-year agreement with the French company Mistral AI covering all of Mistral's commercial models across the group's business lines, after a first use in Global Markets from 2023. In May 2026 the bank extended the partnership for three years, now including software and joint research projects in addition to model access.
Computing power in Germany
Sovereign models need hardware under the same law. Deutsche Telekom's T-Systems runs an AI factory with close to 10,000 Nvidia Blackwell GPUs in Munich's Tucherpark, in a data center operated by Polarise, as DatacenterDynamics reports. The SOOFI project's open European language model of about 100 billion parameters is hosted there.
At the European level, the EU's AI gigafactory program funds larger sites. For a bank the hardware question also touches the energy rules for data centers in Germany.
Upcoming events on digital infrastructure
What is sovereign AI?
Sovereign AI is AI that its user controls: the data stays under the law the user chose, and the model can be run and changed without depending on one foreign provider. Countries use the term for national compute and models, banks for their own platforms.
Does the EU AI Act require sovereign AI?
No. The EU AI Act sets duties by risk class, such as for credit scoring, and applies to providers and deployers wherever the model runs. Sovereignty is a choice a bank makes for control. DORA's rules on ICT third-party risk apply to a model provider as to any other ICT provider.
Is sovereign AI the same as on-premise AI?
Not necessarily. A model on the bank's own servers is one form. A model in a sovereign cloud, operated by an EU company under EU law, is another. What counts is who can access the data and who can stop the service.
Sovereign AI and Finance Loop
Finance Loop lists the AI conferences where banks discuss sovereignty, such as KI Exchange 2026 in Hamburg, with sovereign AI infrastructure as a main theme, and fAInance 2026 in Frankfurt, with a keynote on agentic AI and digital sovereignty. Finance Loop connects the AI leads of banks with model and infrastructure providers in its Digital Infrastructure & Sovereignty track, and membership opens that network.
Finance Loop is a professional network and has the goal of driving the adoption of emerging technologies in finance, such as AI, digital payments, cloud and blockchain solutions. Finance Loop helps its members build skills and personal networks in these fields: Investment & Digital Assets, Payments & Digital Money, Digital Infrastructure & Sovereignty, and Risk & Compliance.