NIST AI Risk Management Framework — Generative AI Profile (NIST AI 600-1)

Adopt with guardrails

Why

It is the most widely cited, US-government-anchored generative-AI risk taxonomy and is already surfacing as a de facto procurement/underwriting reference point in the ecosystem pragma.vision's own AI agents will be evaluated against; adopting its vocabulary keeps our risk language interoperable without over-claiming a binding status the document itself does not have (voluntary, no audit body, EO anchor rescinded, and the strongest 'industry standard' claims trace to a single commercially-interested vendor rather than independent confirmation).

Next

Reuse NIST AI 600-1's 12-risk taxonomy (CBRN uplift, confabulation/hallucination, harmful bias & homogenization, dangerous/violent/hateful content, data privacy, human-AI configuration, information integrity, information security, intellectual property, obscene/degrading content, value-chain/component integration, environmental) as the reference vocabulary for Rule #26 North Star AI-risk scoring and BIL gate definitions -- cite it as a vocabulary source, never claim it as a certification or compliance pass.