AI Governance Tools Field reference
Chapter 04 · Guard

AI guardrails

Short answer · what are ai guardrails

This chapter lists 3 evidence-gated tools (Azure AI Content Safety, Lakera, NVIDIA NeMo Guardrails), backed by 131 single-source dated facts that cleared deterministic and two-judge review.

§4.1 Listing

Guardrail tools and frameworks

Table 4.1Guardrail tools and frameworks
#ToolLicenceWhat it doesEvidenceLast fact
01Azure AI Content SafetyLicence unknownIn 2023, TechCrunch reported that Azure AI Content Safety was a productized version of the safety system powering Microsoft's chatbot in Bing and GitHub's Copilot code-generating service.
02LakeraLicence unknownIn 2025, Atomico reported that Lakera’s co-founders were David Haber, Matthias Kraft and Mateo Rojas-Carulla.
03NVIDIA NeMo GuardrailsLicence unknownAppSecSanta describes NeMo Guardrails as an open-source toolkit by NVIDIA for adding programmable guardrails to LLM-based conversational applications.
§4.2 Placement

Where a guard sits: input, output, tool call

Request path with an input guard before the model, an output guard after it, and a tool-call guard on the branch to external tools. PROMPT INPUT GUARD MODEL OUTPUT GUARD REPLY TOOL-CALL GUARD TOOL
Fig. 4.2Three placements. Each guard is a filter on one edge of the path, not a property of the model.
§4.3 Evidence

Gated evidence highlights

Azure AI Content Safety

In 2023, TechCrunch reported that Azure AI Content Safety was a productized version of the safety system powering Microsoft's chatbot in Bing and GitHub's Copilot code-generating service.

  • In 2023, TechCrunch reported that Azure AI Content Safety was a productized version of the safety system powering Microsoft's chatbot in Bing and GitHub's Copilot code-generating service. (techcrunch.com, verified )
  • In 2023, SCHNEIDER IT MANAGEMENT reported that Microsoft Azure AI Content Safety was an AI-powered content moderation platform. (schneider.im, verified )
  • Content Safety in Foundry Control Plane exposes Text and Image APIs and offers a Studio experience. (azure.microsoft.com, verified )
  • Guardrails are available in Microsoft Foundry for models sold by Azure and models from partners and the community, or as a standalone offering through Azure AI Content Safety Studio and APIs. (learn.microsoft.com, verified )
  • Azure AI Content Safety provides an API/SDK for integrating content safety checks into applications or platforms. (techcommunity.microsoft.com, verified )
  • Content Safety Studio includes Microsoft's built-in terms blocklists and supports uploading custom blocklists. (learn.microsoft.com, verified )

Lakera

In 2025, Atomico reported that Lakera’s co-founders were David Haber, Matthias Kraft and Mateo Rojas-Carulla.

  • In 2025, Atomico reported that Lakera’s co-founders were David Haber, Matthias Kraft and Mateo Rojas-Carulla. (atomico.com, verified )
  • In 2023, TechCrunch reported that Lakera was founded out of Zurich in 2021. (techcrunch.com, verified )
  • In 2025, GlobeNewswire reported that Lakera was dual-headquartered in Zurich and San Francisco. (globenewswire.com, verified )
  • In 2025, Globes reported that Lakera had its headquarters in Zurich. (en.globes.co.il, verified )
  • Lakera’s trust center states that Lakera is part of Check Point Software Technologies. (trust.lakera.ai, verified )
  • In 2025, Globes reported that Check Point Software Technologies Ltd. had entered into an agreement to acquire Lakera. (en.globes.co.il, verified )

NVIDIA NeMo Guardrails

AppSecSanta describes NeMo Guardrails as an open-source toolkit by NVIDIA for adding programmable guardrails to LLM-based conversational applications.

  • AppSecSanta describes NeMo Guardrails as an open-source toolkit by NVIDIA for adding programmable guardrails to LLM-based conversational applications. (appsecsanta.com, verified )
  • In 2025, SiliconANGLE reported that NeMo Guardrails is a protective framework for developers that moderates AI models. (siliconangle.com, verified )
  • AppSecSanta says NeMo Guardrails ships as a NeMo Microservice inside the NVIDIA AI Enterprise software suite for NVIDIA-native deployments. (appsecsanta.com, verified )
  • In 2025, NVIDIA reported that NeMo Guardrails defines and orchestrates guardrails and enforces them for content safety and jailbreak prevention in AI agents and other generative AI applications. (developer.nvidia.com, verified )
  • NVIDIA documentation says configurations are portable between the NeMo Guardrails library and microservice, supporting local development with the library and production deployment with the microservice. (docs.nvidia.com, verified )
  • NVIDIA documentation lists a Python SDK, API server, hooks server, and gateway server as runtime interfaces connecting checks to an application or agent. (docs.nvidia.com, verified )
§4.4 Citations

Sources

  1. ai.azure.com/catalog/models/Azure-AI-Content-Safety, accessed
  2. appsecsanta.com/nemo-guardrails, accessed
  3. atomico.com/insights/lakera-acquisition-making-generative-ai-safer-for-the-world, accessed
  4. azure.microsoft.com/en-us/pricing/details/content-safety, accessed
  5. azure.microsoft.com/en-us/products/ai-services/ai-content-safety, accessed
  6. developer.nvidia.com/blog/defining-llm-red-teaming, accessed
  7. docs.lakera.ai/docs/landing, accessed
  8. docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/overview, accessed
  9. en.globes.co.il/en/article-check-point-buys-swiss-ai-security-co-lakera-1001521951, accessed
  10. github.com/NVIDIA-NeMo/Guardrails/security, accessed
  11. learn.microsoft.com/en-us/azure/ai-services/content-safety/overview, accessed
  12. learn.microsoft.com/en-us/azure/foundry/responsible-ai/content-safety/transparency-note, accessed
  13. platform.lakera.ai/pricing, accessed
  14. siliconangle.com/2025/01/16/nvidia-releases-microservices-safeguard-ai-agents, accessed
  15. techcommunity.microsoft.com/blog/azure-ai-foundry-blog/announcing-general-availability-of-azure-ai-content-safety/3956309, accessed
  16. techcrunch.com/2023/05/23/microsoft-launches-new-ai-tool-to-moderate-text-and-images, accessed
  17. techcrunch.com/2023/10/12/lakera-launches-to-protect-large-language-models-from-malicious-prompts, accessed
  18. trust.lakera.ai, accessed
  19. checkpoint.com/press-releases/check-point-acquires-lakera-to-deliver-end-to-end-ai-security-for-enterprises, accessed
  20. cnbc.com/2023/04/25/nvidia-nemo-guardrails-software-stops-ai-chatbots-from-hallucinating.html, accessed
  21. globenewswire.com/news-release/2025/09/16/3150869/0/en/Check-Point-Acquires-Lakera-to-Deliver-End-to-End-AI-Security-for-Enterprises.html, accessed
  22. lakera.ai, accessed
  23. lakera.ai/ai-red-teaming, accessed
  24. lakera.ai/customers, accessed
  25. schneider.im/microsoft-azure-ai-content-safety-available, accessed
  26. sec.gov/Archives/edgar/data/1015922/000117891325003335/exhibit_99-1.htm, accessed
  27. sec.gov/Archives/edgar/data/1015922/000117891325003986/R22.htm, accessed
  28. securityweek.com/check-point-to-acquire-ai-security-firm-lakera, accessed
  29. sysdig.com/blog/securing-nvidia-ai-stacks-for-enterprise-environments, accessed