ETSI ISG SAI has released three key reports to address AI security, privacy, and transparency for the standards community. ETSI GR SAI 007 outlines steps for designers to ensure AI systems are explainable and transparent, helping to identify biases such as trait-based filtering in decision-making engines. ETSI GR SAI 009 establishes a security framework for AI computing platforms to protect valuable assets like models and data through specific hardware and software mechanisms. Finally, ETSI GR SAI 013 introduces a lightweight framework for multi-partner proofs of concept to demonstrate the applicability of these security solutions and raise awareness of practical AI threats.
Keywords: AI security framework, explicability and transparency, privacy by default, multi-partner PoC, AI threat mitigation