NIST says AI 800-3 gives evaluators a clearer statistical framework by separating benchmark accuracy from generalized accuracy and by introducing generalized linear mixed models for uncertainty estimation. The February 19, 2026 report argues that many current benchmark comparisons hide assumptions that can distort procurement, development, and policy decisions.
NIST on March 9, 2026 published NIST AI 800-4, a report on the challenges of monitoring deployed AI systems. It organizes post-deployment AI oversight into six categories spanning functionality, operations, human factors, security, compliance, and large-scale impacts.
NIST’s CAISI released draft guidance NIST AI 800-2 for automated language-model benchmark evaluations and opened comments through March 31, 2026. The draft focuses on objective setting, execution methodology, and analysis/reporting quality.