From Models to Systems: The Evolution of T&E for Artificial Intelligence from 2020 to 2030

Test and evaluation (T&E) for artificial intelligence is entering a new phase. In 2020, the central challenge was to treat machine learning models as system functions rather than as ordinary software components. By 2025, that view had expanded: AI‐enabled systems had to be evaluated in operational, organizational, and lifecycle context, with attention to trust, adaptation, resilience, and human‐AI teaming. Looking toward 2030, an even larger shift is emerging. AI is no longer only part of the delivered capability; it is increasingly part of the engineering machinery that produces requirements, code, tests, models, documentation, and assurance evidence. This article traces that progression from model‐centric assurance to lifecycle‐centered assurance and, ultimately, to T&E for AI‐generated systems. It argues that future assurance will depend not only on testing delivered systems, but also on evaluating the provenance, governance, and stability of the generative pipelines that create and modify them.