A voluntary framework organised around four functions — Govern, Map, Measure and Manage — intended to help organisations address risks throughout the AI lifecycle. It carries no legal force of its own.
Its significance is indirect and growing: several US state statutes treat documented alignment with a recognised framework as evidence of reasonable care, and the AI RMF is the framework they have in mind. That converts a voluntary standard into something with legal consequence, and raises the bar on what "documented alignment" has to mean in practice.