AI agents have the potential to transform the productivity of billions of knowledge workers around the world through AI devices that perform a variety of tasks. However, companies must address critical issues such as trust, safety, security, and compliance to develop AI agents.

NVIDIA announced today the launch of the NVIDIA NIM microservice for AI guardrails. The NIM microservice reportedly enables developers to build secure and reliable AI agents.

The new NVIDIA NIM microservice for AI Guardrail is part of NVIDIA NeMo Guardrail, a suite of software tools that are portable, optimized inference microservices that help enterprises improve the safety, accuracy, and scalability of generative AI applications.

Nemo Guardrail helps curate, customize, and guardrail AI. Nemo Guardrail helps developers integrate and manage AI guardrails into their large language model (LLM) applications. △Amdocs △Cerence AI △Rose, etc. are using Nemo Guardrail to protect AI applications.

NVIDIA introduced three new NIM microservices for Nemo Guardrails that help AI agents maintain controlled behavior while operating at scale.

The Content Safety NIM microservice ensures that AI responds ethically by preventing biased or harmful outputs. The Topic Control NIM microservice keeps conversations focused on approved topics, avoiding off-topic or inappropriate content. The Jailbreak Detection NIM microservice adds protection against jailbreak attempts, helping maintain AI integrity in adversarial scenarios.

By applying a variety of lightweight specialized models as guardrails, developers can fill in gaps that cannot be blocked by general global policies and protections alone. Because a one-size-fits-all approach cannot adequately secure and control complex agentic AI workflows.

“Technologies like Nemo Guardrail are essential to securing generative AI applications and ensuring they operate safely and ethically,” said Anthony Gunettilek, Group President of Technology at Amdocs. “By integrating it into our platform, we are strengthening our ‘Trusted AI’ capabilities.”

“We use NVIDIA’s high-performance security solutions for our in-car assistant technology,” said Niels Schantz, vice president of technology at Serence. “Nemo Guardrail enables us to deliver illusion-free responses, is custom-built for our automaker customers to filter out harmful or offensive requests, and our CaLLM family of language models protects against delivering unintended or inappropriate content to end users.”