SAFIR: Secure AI For Intelligent Resilience and Confidentiality in the Cloud

Purpose and goal

The SAFIR project aims to build secure, resilient and trustworthy cloud-edge infrastructures for AI-driven workloads by combining confidential computing, secure high-speed data flows and AI-based orchestration. Its objectives are to ensure trusted execution of sensitive AI applications, protect data in transit and at rest, and enable safe integration of AI into cloud control systems while anticipating societal events that impact digital demand.

Expected effects and result

SAFIR is expected to deliver new methods and tools for confidential AI execution, secure RDMA communication and verified AI-driven orchestration. The project will improve cybersecurity, resilience and performance of cloud-edge systems, reduce risks of data leakage and service disruption, and enable proactive, socio-aware resource management. These results will strengthen Sweden’s competitiveness in secure digital infrastructure and accelerate trusted AI adoption across critical sectors.

Planned approach and implementation

The project is structured around two main tasks: developing trustworthy cloud-edge infrastructures and advancing secure AI for self-adapting systems. It combines research and industrial collaboration to integrate TEEs, secure high-speed communication, AI-based verification and socio-aware orchestration. Implementation includes iterative development, validation in realistic environments, and deployment of demonstrators, ensuring solutions are scalable, secure and applicable to industry needs.