NIST asks feedback on its TEVV-Athlon framework to verify AI systems achieve goals & minimize harms

NIST seeks feedback on its TEVV-Athlon Framework, a flexible, four-stage methodology developed to help organizations design customized Test, Evaluation, Verification, and Validation (TEVV) assessments tailored to the real-world risks and objectives of their AI systems. Adaptable across diverse technologies, including LLMs, agentic systems, and multi-modal models, it structures evaluations around specific Events and Tools that capture performance data across targeted measurement domains (Blocks). By producing context-driven, repeatable evidence, the framework can verify that an AI system reliably achieves its intended goals while minimizing potential harms. Read the complete TEVV-Athlon Framework here.