What TEVV covers
TEVV stands for testing, evaluation, verification and validation. NIST describes a four-stage approach for constructing assessments around organisational objectives, rather than assuming one test suits every AI application. [1]
The current publication status
The announcement identifies NIST AI 200-2 as an initial public draft. Its stated comment period ended on 6 October 2026. A closed comment period does not by itself mean a final standard has been issued or a product has been approved. [1]
ADM perspective
For a scheduling tool, assess missed constraints, operational disruption and recovery from bad input. For a document assistant, assess unsupported statements, useful coverage and escalation. Report what was tested, with which data and controls, before comparing scores. Verification checks specified requirements; validation asks whether the system meets the intended need. Those questions can lead to different evidence.
Read the original sources
The original publications provide the full detail, evidence and limitations.
- NIST · Initial public draft announcementThe TEVV-Athlon Framework for Evaluating AI Systems
An AI-assisted, source-led summary. “ADM perspective” is editorial interpretation, separate from the source’s findings. Our editorial approach.