Keeping clinical evidence current as medical artificial intelligence systems change

Generative and agentic artificial intelligence systems may change faster than clinical evidence can be generated, reviewed, and implemented. Updates to the foundation model, prompt, retrieval layer, generation settings, tools, memory, interface, or workflow can alter behaviour without changing the product name visible to investigators or users. This Perspective synthesizes evidence on repeated-run instability, cross-version performance, human–AI interaction, agentic systems, and post-deployment monitoring. We propose VERSION-AI as a testable evidence-refresh framework that separates what changed from the clinical consequences of failure. C0–C8 describe the system or context change, whereas a distinct clinical-consequence tier reflects intended use, decision dependence, time sensitivity, and the likelihood and severity of harm from incorrect, omitted, or delayed actions. The combination determines whether evidence may remain current after audit, requires usability or repeated-run bridging, cross-version or local validation, shadow evaluation, continuous monitoring, rollback capability, or renewed prospective clinical assessment. Because VERSION-AI has not yet been prospectively validated, it is presented as a research and governance proposal rather than a regulatory instrument. Evidence transfer after a material system change should be demonstrated rather than assumed.