Nemotron Policy Generator
Build content-safety policies for NCS, NCS-VL, NCS-Reasoning, or NeMo Guardrails
Load Example
Reset
1. Policy Metadata
Policy Name
Version
Owner
Date
2. Target Model & Use Case
Target Nemotron model
NCS (text-only)
NCS-VL (vision-language)
NCS-Reasoning (CoT-aware)
NeMo Guardrails wrapper
Deployment archetype
Consumer chatbot
Enterprise RAG / copilot
Kids / education
Healthcare / clinical
Financial services
Code assistant / developer tools
Government / sovereign
Use cases (select all that apply)
Runtime guardrails
Eval rubric
Training-data labeling
Customer BYO-policy
Severity model
Graded S0–S4 (recommended for runtime)
Binary (block / allow)
Jurisdiction
3. Categories
Toggle V2 canonical categories; add custom categories below.
V2 canonical
Custom categories
+ Add custom category
4. Allow-list (explicit affordances)
What's explicitly permitted even if it sounds adjacent to a blocked category. One per line.
5. Calibration & Response
Calibration notes
Response guidance (one line per severity)
S0
S1
S2
S3
S4
6. NCS-Reasoning rules (if applicable)
S3+ categories trigger on the CoT alone — unsafe reasoning with a safe final answer is still a violation. S1–S2 are evaluated on the final answer only. The allow-list applies to both CoT and answer.
7. Export
Download Markdown
Download JSON
Download System Prompt
Download All (zip-less)
Markdown
JSON
Nemotron System Prompt
Copy to clipboard
Rebuild preview