What happened
The 60-day public comment period on the initial public draft (ipd) of NIST AI 200-2, 'The TEVV-Athlon Framework for Evaluating AI Systems,' closed on (or about) 6 October 2026. The draft was first announced by NIST on 7 August 2026 and defines a four-stage structured method (Articulate & Organize, Define Measurement Concepts, Construct Activities, Synthesize & Report) for customizing Test, Evaluation, Verification, and Validation (TEVV) assessments of AI systems, explicitly covering statistical ML models, LLMs, multimodal models, and agentic systems. It is positioned as the implementation of the NIST AI RMF's Measure function. Multiple independent sources (NIST comment window tracker, practitioner public-comment submissions dated 9 Sep 2026, and analyst write-ups) confirm the draft and its in-window closure.
Why it matters
TEVV-Athlon is NIST's central delivery vehicle for the AI RMF's Measure function and will anchor how US federal agencies and contractors plan and defend AI/agent evaluation evidence. Its comment closure marks the point at which NIST consolidates feedback on the first framework built to evaluate agentic AI security behavior — the subject of the entire 2026 agent-security guidance wave. Comments received (e.g. on operationalizing interruption/kill-switch capability and evidence custody) will shape the final edition.
Action needed
Track the second public draft/final AI 200-2 release; map existing AI evaluation playbooks to the four TEVV-Athlon stages and review the comment summary when NIST publishes it.