NIST AI Risk Management Framework

The NIST AI RMF is voluntary — but through federal procurement and industry practice it has become the de facto reference for "trustworthy AI" in the United States. Its measurement functions are where the Causal Seal fits.

The four functions

The AI RMF (version 1.0) organizes trustworthy-AI practice into four functions: Govern (culture and accountability), Map (context and risk framing), Measure (analyze, track, and document system behavior), and Manage (act on prioritized risks). Two of the framework's characteristics of trustworthy AI — Accountable & Transparent and Explainable & Interpretable — depend directly on the ability to evidence why a system behaved as it did.

Where the seal contributes

RMF function / characteristic	Causal Seal contribution
MEASURE — track and document AI system behavior over time	A per-output, tamper-evident record of the causal state — a continuous, machine-verifiable measurement substrate rather than sampled logging.
MEASURE / MANAGE — traceability of decisions for incident review	Any past generation is replayable from its seal; corrective and rupture events are individually flagged and filterable.
GOVERN — accountability & auditability	Optional Ed25519 signatures bind each decision to a named emitter, making responsibility assignable.
Explainable & Interpretable	The parameter dictionary publishes the meaning of every sealed field, so an auditor reads a decision in human terms — not in model internals.

The AI RMF is a framework, not a checklist of pass/fail controls. The Causal Seal does not "achieve RMF compliance"; it provides concrete, verifiable evidence that supports the Measure and Manage functions and the transparency characteristics — evidence an organization can point to in its own risk documentation.

→ US state regimes (Colorado & the patchwork) · → EU AI Act