Creating Privacy Preserving AI with Substra — Federated Learning Framework

AI Impact Summary

Substra offers a solution to the challenge of training AI models with sensitive data by enabling Federated Learning. This approach allows models to be trained across distributed datasets without directly sharing the underlying data, addressing concerns around data privacy and compliance with regulations like HIPAA. The framework’s real-world deployments, such as in breast cancer research, demonstrate its potential to unlock valuable insights from siloed data sources, while also mitigating bias inherent in single-source datasets.