NIST’s AI Risk Management Framework organises responsible practice around four connected functions: GOVERN, MAP, MEASURE and MANAGE. Its significance lies in refusing to treat risk as a single model score. An organisation must establish who governs, delimit contexts and affected populations, choose evaluative instruments and act upon their findings. The functions are iterative, so measurement is not meaningful without capacities to revise policy or implementation. This creates a useful architecture for tracing a decision from design assumptions to downstream harm, including cases where technical validity coexists with adverse institutional consequences. The framework is voluntary guidance rather than law; its existence does not demonstrate that an organisation complies with human-rights duties or offers usable remedies to affected individuals. Methodologically, it bridges engineering risk assessment, organisational governance and social accountability. In Socioplastics, it invites a diagnostic reversal: an archive or decision infrastructure should not be assessed only by how well it represents a problem, but also by how it handles the evidence that its own categories fail.