United States of America: National Institute of Standards and Technology issued report on challenges related to monitoring of deployed AI systems

On 9 March 2026, the National Institute of Standards and Technology (NIST) published a report (NIST AI 800-4) examining challenges in monitoring deployed AI systems. The report, based on a year-long inquiry led by the Center for AI Standards and Innovation (CAISI) with input from model developers, compute providers, and federal agencies, outlines a framework for post-deployment measurement to ensure AI systems operate reliably, track unforeseen outputs, and identify unexpected outcomes in changing contexts. It identifies six key monitoring categories, functionality, operational consistency, human factors, security, regulatory compliance, and large-scale impacts, and classifies challenges into cross-cutting issues, category-specific challenges, and open questions, such as optimal monitoring frequency and balancing automated versus human oversight. While relevant across the AI ecosystem, the report highlights particular importance for frontier generative AI systems and provides a foundation for future guidance on field studies and incident monitoring.