Intel's Secure Data Tunnel Moves AI Training Models to Data Sources

The article describes Intel's Tiber Secure Federated AI service, which uses confidential computing and on‑chip TDX mechanisms to enable federated model training where models travel to data rather than moving data offsite. It outlines target use cases (e.g., healthcare and finance), practical challenges around data quality and collaboration, and highlights security concerns by referencing disclosed vulnerabilities and prior research that broke into confidential computing implementations, leading experts to question current foundational approaches.