NVIDIA (CEO Jensen Huang), a leader in AI computing technology, has released open-source generative AI software that adds safeguards for AI chatbots, helping to securely protect a new type of AI-based application.

NVIDIA announced on the 26th that it has released NVIDIA NeMo Guardrails, open source software that supports the generation of impressive, high-quality text responses for generative AI applications.

Nemo Guardrail helps verify the accuracy, appropriateness, relevance, and safety of smart applications powered by Large Language Models (LLM). Additionally, Nemo Guardrail includes all the code, examples, and documentation necessary for enterprises to add safety to text generation AI apps.

The launch of Nemo Guardrail reflects the widespread adoption of LLM, a powerful engine for AI apps, across the industry. LLM is being utilized extensively for everything from responding to customer inquiries and summarizing long documents to writing software and accelerating new drug design.

Square guardrails are designed to allow users to safely protect these new types of AI-based applications.

The safety of generative AI is a concern across the industry. NVIDIA's Nemo Guardrail is designed to work with all LLMs, such as OpenAI's ChatGPT. Through Square Guardrails, developers can safely coordinate LLM-based apps and keep them within the company's area of expertise.

Developers can set three types of boundaries using square guardrails.

First, topical guardrails prevent the app from veering into unwanted areas. For example, they prevent a customer service assistant from answering questions about the weather.

Safety guardrails ensure that apps respond with accurate and relevant information. They can filter out unwanted language and force the mention of only trusted sources.

Security guardrails restrict apps to connecting only to external third-party applications known to be safe.

If you use Square Guardrails, almost any software developer, even those who are not machine learning experts or data scientists, can quickly generate new rules with just a few lines of code.

Since Nemo Guardrail is open source, it can work with all tools used by enterprise app developers.

For example, Nemo Guardrail can run on LangChain, an open source toolkit that developers are rapidly adopting to connect third-party applications to LLM.

In addition, Nemo Guardrail is designed to work with a wide range of LLM-supported applications such as Zapier. Zapier is an automation platform used by over 2 million companies, and it has seen firsthand how users integrate AI into their work.

NVIDIA is integrating the Nemo Guardrail into the NVIDIA Nemo Framework. The NVIDIA Nemo framework includes everything users need to train and tune language models using the company's proprietary data.

Most of the Nemo framework is already available as open source code on GitHub. For enterprises, it can also be obtained as a complete support package as part of the NVIDIA AI Enterprise software platform.


Nemo is also available as a service and is part of NVIDIA AI Foundation, a suite of cloud services for enterprises looking to build and run custom generative AI models based on their own datasets and domain knowledge.