New Federated Data Network Enables Private AI Training
The platform allows clinical researchers to train models across institutions without moving sensitive patient data.
Updated on Oct. 27, 2026 in Artificial Intelligence
Live Poll
Should healthcare providers allow patient data to be used for secure medical research?
Apheris and T-Systems have deployed a federated data network for clinical research, utilizing the Industrial AI Cloud to train models across multiple European hospitals without transferring sensitive patient records. This infrastructure currently supports a network of over 50,000 patients.
Why it matters
The implementation addresses data sovereignty constraints under the General Data Protection Regulation, which previously complicated multi-site clinical studies. By enabling collaborative research without centralizing sensitive medical data, the platform accelerates large-scale drug discovery.
The platform utilizes NVIDIA FLARE as its core federated learning framework, enabling model parameters to be trained locally at medical sites. These parameters are then aggregated within T-Systems' Industrial AI Cloud, serving a network that currently includes 9 pharmaceutical companies.
The players
Apheris
A software company that specializes in federated data networks and privacy-preserving AI for drug discovery and clinical research.
T-Systems
A subsidiary of Deutsche Telekom that provides ICT solutions, including the Industrial AI Cloud infrastructure for GPU-heavy workloads.
The details
Federated learning works by moving the model to the data rather than the data to the model, ensuring that patient information remains within the local environment of the clinical institution. During training, individual hospital systems calculate updates based on their local MRI and healthcare data, transmitting only these mathematical parameters to the cloud. A central server then performs global model aggregation, incorporating insights from all participating institutions while maintaining strictly localized data privacy.
Timeline
A feasibility study on MS lesion segmentation was completed in 2025.
A presentation regarding the partnership is scheduled for October 20 to 22, 2026.
The Tech Race
This deployment marks a shift in how pharmaceutical companies approach large-scale structural biology research by avoiding the prohibitive legal overhead of data centralization. It directly competes with siloed clinical research models by offering a standardized, privacy-first infrastructure.
Pharmaceutical and clinical researchers can now initiate large-scale multi-site studies without waiting for the complex data-sharing agreements typically required for centralized databases. The system currently supports a network of nine pharmaceutical companies and over 50,000 patients.
The takeaway
Federated learning provides a path forward for secure, multi-institutional clinical research that satisfies stringent European data privacy laws. Researchers should track the upcoming presentation at the NVIDIA GPU Technology Conference in October 2026 for further insights into the framework's scalability.
What happens next
The partnership details will be presented at the NVIDIA GPU Technology Conference held in Berlin from October 20 to 22, 2026.
Further reading
Explore more developments in Artificial Intelligence as research organizations scale privacy-preserving compute.
Source note: This article includes information reported by TelecomDrive.
Live Poll
Should healthcare providers allow patient data to be used for secure medical research?
