What happened
ETSI published Technical Specification TS 104 033 on June 2, 2026, establishing a systematic security baseline for AI computing platforms — the infrastructure layer that hosts AI model training and inference. The specification defines security requirements and functions for AI computing platforms, specifying required security components and their interfaces, and is designed to support compliance with the EU AI Act. It was developed by ETSI's Technical Committee Securing AI (TC SAI).
Why it matters
TS 104 033 is the first formal European technical standard addressing security at the AI compute layer rather than at the application or output level. By explicitly targeting compliance with the EU AI Act and addressing threats to models and data at the platform level, it establishes a procurement and audit baseline for AI infrastructure providers, cloud operators, and enterprises deploying on-premise AI compute. Auditors and enterprise security architects now have a normative reference for AI platform security requirements.
Action needed
AI platform architects and cloud security teams should obtain TS 104 033 and map existing AI compute environments against its requirements. Consulting teams supporting EU AI Act compliance programs should add TS 104 033 to the evidence baseline for Article 15 (cybersecurity) and technical documentation requirements.