NeMo Guardrails

• Publisher: NVIDIA
• Status: active
• Version: 0.15.0
• Release Date: 2025-08-08
• Date Added: 2025-08-25
• Source URL: https://github.com/NVIDIA/NeMo-Guardrails

---

Summary

Open-source toolkit for securing LLM-powered apps and AI agents. NeMo Guardrails enforces "rails" that filter inputs/outputs, identify jailbreaks and prompt injection attempts, manage conversation state, and fact-check answers. Useful for stopping data leaks, preventing rogue agent actions, and providing audit trails for security teams.

---

Key Takeaways

• Block unsafe prompts before they reach the model or agent.
• Catch jailbreaks and prompt injection attempts.
• Stop unsafe outputs — detect hallucinations and fact-check results.
• Write custom guardrails (Colang) for rules like PII masking or restricted actions.
• Log and trace activity for audits, monitoring, and incident response.
• Integrates with LangChain (agent framework).

---

Related Code

/code/security-tools/nemo-guardrails

---

Additional Sources

• arXiv Paper — NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Rails
• NVIDIA NeMo Guardrails Documentation

---

Tags

guardrails, filtering, llm-security, agentic-ai, monitoring, jailbreak, prompt-injection, fact-checking, audit-trails, safety, alignment

---

License

Apache-2.0