A profile, not a new framework
NIST describes AI 600‑1 as a cross-sectoral profile of AI RMF 1.0 and a companion resource to it. The four functions and their subcategories stay the same.
NIST AI RMFNIST AI 600‑1 explained: the twelve generative AI risks it defines, how its suggested actions map to the AI RMF and where the profile stands today.
The NIST Generative AI Profile is NIST AI 600‑1, a companion to the NIST AI Risk Management Framework that applies the framework to generative AI.
NIST released it on July 26, 2024. It defines twelve risks that are unique to or exacerbated by generative AI and suggests actions for managing them, organized by AI RMF subcategory. Use is voluntary.
risks unique to or exacerbated by generative AI, defined in section 2 of the profile
Source: NIST AI 600‑1, July 2024
primary considerations the public working group focused on, from governance to incident disclosure
Source: NIST AI 600‑1, July 2024
the date NIST released the profile, eighteen months after AI RMF 1.0
Source: NIST, AI RMF overview
Section 2 of the profile defines twelve risks. Each is summarized here in a line. The profile gives the full definitions.
| Risk | What NIST means by it |
|---|---|
| CBRN information or capabilities | Easier access to information or design capabilities for chemical, biological, radiological or nuclear weapons. |
| Confabulation | Confidently stated but false content, known colloquially as hallucinations or fabrications. |
| Dangerous, violent or hateful content | Easier production of violent, inciting, radicalizing or threatening content. |
| Data privacy | Leakage, unauthorized use or de-anonymization of personal and sensitive data. |
| Environmental impacts | High compute resource use in training or operating generative AI models. |
| Harmful bias or homogenization | Amplified societal and systemic biases, and undesired sameness in outputs. |
| Human-AI configuration | How people and AI interact, including automation bias, over-reliance and emotional entanglement. |
| Information integrity | A lower barrier to spreading content that may not separate fact from opinion or fiction. |
| Information security | Lower barriers for offensive cyber capabilities, and a larger attack surface around training data, code and model weights. |
| Intellectual property | Easier replication of allegedly copyrighted, trademarked or licensed content without authorization. |
| Obscene, degrading or abusive content | Easier production of obscene, degrading or abusive imagery. |
| Value chain and component integration | Non-transparent or untraceable upstream third-party components, and improper supplier vetting. |
The profile does not replace the framework. It fills in the framework’s structure for one technology.
NIST describes AI 600‑1 as a cross-sectoral profile of AI RMF 1.0 and a companion resource to it. The four functions and their subcategories stay the same.
NIST AI RMFSuggested actions sit under GOVERN, MAP, MEASURE and MANAGE subcategories. Each has an ID, such as GV‑1.1‑001 for the first action under GOVERN 1.1, and the risks it addresses.
The public working group behind the profile focused on four primary considerations: governance, content provenance, pre-deployment testing and incident disclosure.
Read NIST AI 600‑1Treat the profile as a menu matched to your own systems, not a list to complete from top to bottom.
Inventory the generative AI you build, buy and embed, including models and components from third parties.
Go through the twelve risks for each system. A support chatbot and a code assistant will not share a list.
Use the risk tags to find the suggested actions for those risks under each subcategory.
Give each action an owner and keep evidence of what was done for which system.
Repeat when a model, vendor or use case changes, or when NIST revises the profile.
Straight answers, kept current. Last updated .
Ask us anything →No. The profile is a companion resource to the NIST AI Risk Management Framework, which NIST released for voluntary use. It suggests actions and sets no requirements. Organizations use it to bring a recognized structure to generative AI risk.
As of November 17, 2026, NIST’s publication record lists the July 2024 edition with no superseding version. The profile says future revisions will add subcategories, risks and suggested actions. Separately, NIST states that AI RMF 1.0 is being revised as part of the White House AI Action Plan.
Yes. One of the twelve risks is value chain and component integration: non-transparent or untraceable integration of upstream third-party components, including improper supplier vetting. Managing it starts with knowing which models, datasets and vendors each system depends on, which is what an AI‑BOM records.
Primary sources for the figures and frameworks cited on this page. Last verified .
See how Cranium AI discovers, observes, governs, secures and proves every AI model, agent and vendor in your enterprise.
Cookie Policy · Privacy Policy
Global Privacy Control detected. Opt-out request honored: marketing tags stay off in this browser.