ETSI Publishes TS 104 033 for AI Platform Security, Banning Remote Root Access
Updated
Updated · Help Net Security · Jun 4
ETSI Publishes TS 104 033 for AI Platform Security, Banning Remote Root Access
1 articles · Updated · Help Net Security · Jun 4
Summary
TS 104 033 sets concrete security requirements for AI computing platforms used in data centers and edge environments, covering the infrastructure that hosts AI models, datasets, training and inference workloads.
The specification groups controls around identity and access, data and integrity protection, auditing, incident response and resilience, and explicitly prohibits remote access to root-level accounts while requiring secure boot and protected data transmission.
AI-specific provisions target shared accelerators such as GPUs and NPUs, requiring tenant isolation, inference-attack detection, secure AI logging, training recovery after failures or cyberattacks, and support for Model Bill of Materials records.
ETSI said the standard builds on its earlier AI computing platform security framework and is meant to provide a foundation for trustworthy AI operations across cloud, data center and edge deployments.
Will this new AI security standard inadvertently stifle innovation and favor large tech corporations?
How can organizations realistically detect microscopic data poisoning attacks that can corrupt an entire AI model?
As AI-driven attacks evolve daily, can a static security framework truly protect against future, unknown threats?
Establishing a 72-Requirement Security Baseline: How ETSI TS 104 033 Redefines AI Platform Trust in 2026
Overview
ETSI TS 104 033, released in May 2026, sets a critical security baseline for AI computing platforms by defining comprehensive security criteria for manufacturers and safeguarding essential AI components like core execution environments and service interfaces. This standard addresses the urgent need for robust security in AI, aiming to protect the underlying technology that powers AI systems. By establishing a structured approach, ETSI TS 104 033 ensures a more secure and reliable future for AI development and deployment, marking a significant step forward in building trust and resilience within the rapidly evolving AI landscape.