Welcome Here  Shenzhen Mingjiada Electronics Co., Ltd.

NVIDIA NeMo Guardrails includes new NVIDIA NIM microservices that improve the accuracy, security, and control of AI for organisations building AI across industries.

AI intelligences are expected to become ‘knowledge robots’ capable of completing a wide range of tasks, increasing the productivity of billions of knowledge workers around the world. To develop AI intelligences, organisations need to address key issues such as trust, physical security, cybersecurity and compliance.

The new NVIDIA NIM AI Guardrail microservices are part of the NVIDIA NeMo Guardrails family of software tools. These portable and optimised inference microservices help enterprises improve the security, accuracy and scalability of generative AI applications.

These microservices are orchestrated around the NeMo Guardrails core, which is part of the NVIDIA NeMo Platform for AI management, customisation and guarding, and helps developers integrate and manage AI guardrails in Large Language Model (LLM) applications. Industry leaders such as Amdocs, Cerence AI, and Lowe's use NeMo Guardrails to secure AI applications.

Developers can use NIM microservices to build more secure and trusted AI intelligences. These intelligences will respond securely and appropriately, taking into account context-specific guidelines, and are more resilient to jailbreak attempts. They can be deployed in customer service in industries such as automotive, finance, healthcare, manufacturing, retail, and more to improve customer satisfaction and trust.

One of these new microservices is built specifically for auditing content safety and is trained using the Aegis Content Safety Dataset, one of the higher quality human-annotated data sources of its kind. The dataset is collated and owned by NVIDIA and is publicly available on the Hugging Face website. The dataset contains more than 35,000 samples of manually labelled data that have been tagged for AI security and jailbreak attempts to bypass system restrictions.

NVIDIA NeMo Guardrails Keeps AI Intelligence on Track

AI is rapidly improving the productivity of a wide range of business processes. In transforming customer service, it can speed up the resolution of customer issues by up to 40 per cent. However, in order to scale AI and other AI intelligences for customer service, there is a need for secure models that prevent harmful or inappropriate outputs and ensure that AI applications operate within defined parameters.

NVIDIA has introduced three new NIM microservices for NeMo Guardrails that help AI intelligences operate at scale while maintaining controlled behaviour:

● Content Security NIM microservice: prevents AI from generating biased or harmful output and ensures ethical responses.

● Topic control NIM microservices: focus conversations on vetted topics to avoid off-topic or inappropriate content.

● Jailbreak detection NIM microservices: add protection against jailbreak attempts to help maintain the integrity of the AI in adversarial scenarios.

As a one-size-fits-all approach cannot effectively protect and control complex agent-based AI workflows, by applying multiple lightweight specialised models as guardrails, developers can bridge the gaps that can occur when only more generic global policies and protections are adopted.

Small language models, such as those in the NeMo Guardrails family, have low latency and run efficiently even in resource-limited or distributed environments, making them ideal for scaling AI applications for industries such as healthcare, automotive, and manufacturing, as well as for locations such as hospitals or warehouses.

Industry Leaders and Partners Secure AI with NeMo Guardrails

Now available to the open source community, NeMo Guardrails helps developers improve the security and control of LLM applications by orchestrating multiple AI software policies, known as rails. Together with NVIDIA NIM microservices, it forms a powerful framework for building AI systems that can be deployed at scale without compromising security and performance.

Amdocs, a leading global provider of software and services for media enterprises, is using NeMo Guardrails to enable more secure, accurate, and contextualised answers to enhance AI-driven customer interactions.

Technologies like NeMo Guardrails are critical to securing generative AI applications to help AI operate safely and ethically,’ said Anthony Goonetilleke, president and head of strategy, Amdocs Technology Group. We are enhancing the ‘Trusted AI’ capabilities of our amAIz platform by integrating NVIDIA NeMo Guardrails into the platform to guarantee the security, reliability and scalability of agent-based experiences. This enables service providers to deploy AI solutions safely and securely, setting a new standard for innovation and operational excellence in AI.’

Cerence AI is a company that specialises in AI solutions for the automotive industry. The company is using NVIDIA NeMo Guardrails to help its in-vehicle assistants provide contextually secure interactions supported by its CaLLM family of large and small language models.

Cerence AI relies on NVIDIA's secure, high-performance solutions to power our in-vehicle assistant technology,’ said Nils Schanz, executive vice president of products and technology at Cerence AI. Using NeMo Guardrails helps us provide our automotive customers with a trusted context-aware solution with informed, rigorous and hallucination-free answers. Additionally, NeMo Guardrails can be customised to the needs of our automotive manufacturer customers, helping us to filter harmful or unfriendly requests and preventing our CaLLM family of language models from delivering unexpected or inappropriate content to end users.’

Leading home décor retailer Lowe's is using generative AI to bolster the expertise of its shop associates. By providing more comprehensive product knowledge, these tools enable shop associates to answer customers' questions, help them find the right products for their projects, and set a new standard for retail innovation and customer satisfaction.

We are always looking for ways to help our associates deliver more than what is expected of them to our customers,’ said Chandhu Nair, senior vice president of data, AI and innovation at Lloyd's. With the recent deployment of NVIDIA NeMo Guardrails, we are able to ensure the security and reliability of AI-generated answers, forcing conversations to be limited to relevant and appropriate content.’

To further accelerate the adoption of AI security measures in retail AI application development and deployment, NVIDIA recently announced at the NRF (National Retail Federation) show that NVIDIA AI Blueprint for Retail Shopping Assistants integrates with NeMo Guardrails microservices, creating digital shopping for more reliable and controlled customer interactions for digital shopping.

Consulting leaders Taskus, Tech Mahindra and Wipro are also integrating NeMo Guardrails into their solutions to provide enterprise customers with more secure, reliable and controlled generative AI applications.

With its openness and extensibility, NeMo Guardrails integrates with a robust ecosystem of leading AI security models and protection providers, as well as AI observability and development tools. It supports integration with ActiveFence's ActiveScore to protect against harmful or inappropriate content in conversational AI applications and provides visualisation, analysis and monitoring capabilities.

Hive provides AI-generated content detection models for image, video, and audio content in the form of NIM microservices. By using NeMo Guardrails, the company can easily integrate and orchestrate its models into AI applications.

Fiddler, the AI observability platform, easily integrates with NeMo Guardrails to enhance the monitoring capabilities of AI Guardrails. Weights & Biases, an end-to-end AI developer platform, is building on its existing portfolio of NIM integrations to enhance AI reasoning in production by expanding the capabilities of W&B Weave with the addition of integrations with NeMo Guardrails microservices.