AI agents are poised to transform the productivity of the world’s 1 billion knowledge workers, using “knowledge robots” that can perform a variety of tasks. To develop AI agents, companies must address important concerns such as trust, safety, security, and compliance.
New NVIDIA NIM Microservices for AI Guardrails (part of the NVIDIA NeMo Guardrails collection of software tools) provides portable, optimized inference that helps enterprises improve the safety, accuracy, and scalability of the AI applications they produce. It’s a microservice.
At the heart of microservices orchestration is NeMo Guardrails, part of the NVIDIA NeMo platform for AI curation, customization, and guardrails. NeMo Guardrails helps developers integrate and manage AI guardrails in large-scale language model (LLM) applications. Industry leaders Amdocs, Cerence AI, and Lowe’s are among the companies using NeMo Guardrails to protect their AI applications.
Developers can use NIM microservices to build more secure and reliable AI agents that provide safe and appropriate responses within context-specific guidelines and are hardened against jailbreak attempts. Agents located in customer service across industries such as automotive, finance, healthcare, manufacturing, and retail can increase customer satisfaction and trust.
One of the new microservices built to orchestrate content safety was trained using the Aegis Content Safety Dataset. Aegis Content Safety Dataset is one of the highest quality human annotated data sources in this category. This dataset, managed and owned by NVIDIA, is publicly available on Hugging Face and includes over 35,000 human-annotated data samples flagged for AI safety and jailbreak attempts to circumvent system restrictions. I am.
Keep AI agents on track with NVIDIA NeMo guardrails
AI is rapidly increasing the productivity of a wide range of business processes. Customer service can now resolve customer issues up to 40% faster. However, scaling AI for customer service and other AI agents requires a secure model that prevents harmful or inappropriate output and ensures that AI applications operate within defined parameters.
NVIDIA introduced three new NIM microservices for NeMo Guardrails that enable AI agents to operate at scale while maintaining controlled behavior.
By applying multiple lightweight, specialized models as guardrails, developers can cover gaps that might appear if only more general global policies and protections existed. This is because a one-size-fits-all approach cannot adequately secure and control complex agent AI workflows.
Small language models like those in the NeMo Guardrails collection are designed to have low latency and run efficiently in resource-constrained and distributed environments. This makes it ideal for scaling AI applications in industries like healthcare, automotive, manufacturing, and locations like hospitals and warehouses.
Industry leaders and partners protect AI with NeMo guardrails
NeMo Guardrails, available in the open source community, helps developers orchestrate multiple AI software policies called rails to increase security and control of LLM applications. Works with NVIDIA NIM microservices to provide a robust framework for building AI systems that can be deployed at scale without sacrificing safety or performance.
Amdocs, the world’s leading provider of software and services to telecommunications and media companies, leverages NeMo Guardrails to provide AI-driven customer support by delivering more secure, accurate, and contextually relevant responses. We are enhancing interaction.
“Technology like NeMo Guardrails is essential to protecting generative AI applications and ensuring they operate securely and ethically,” said Anthony Goonetilleke, Group President and Head of Strategy, Technology at Amdocs. states. “Integrating NVIDIA NeMo Guardrails into the amAIz platform strengthens the platform’s Trustworthy AI capabilities and delivers a secure, reliable, and scalable agent experience. This enables service providers to deploy AI solutions securely and confidently, setting a new standard for AI innovation and operational excellence. ”
Cerence AI, a company specializing in AI solutions for the automotive industry, uses NVIDIA NeMo Guardrails to ensure in-vehicle assistants have context-appropriate and safe interactions leveraging the CaLLM family of large and small language models. We support you to provide.
“Cerence AI relies on NVIDIA’s high-performance, secure solutions to power our in-vehicle assistant technology,” said Nils Schantz, executive vice president of products and technology at Cerence AI. “With NeMo Guardrails, we will be able to provide automaker customers with a trusted, context-aware solution to provide smart, thoughtful, and unillusioned responses. It is customizable for customers and helps protect the CaLLM family of language models from unintended or inappropriate content delivery to end users by filtering harmful or objectionable requests.”
Lowe’s, the leading home improvement retailer, leverages generative AI to build on the deep expertise of its store associates. By enhancing access to comprehensive product knowledge, these tools empower employees to answer customer questions, help them find the right product to complete a project, and help retailers setting new standards for innovation and customer satisfaction.
Chandu Nair, senior vice president of data, AI and innovation at Lowe’s, said: “With the recent introduction of NVIDIA NeMo Guardrails, we ensure that AI-generated responses are safe, secure, and reliable, enforcing conversational boundaries and delivering only relevant and appropriate content. ”
To further accelerate the adoption of AI safeguards in AI application development and retail deployments, NVIDIA recently announced at the NRF Show that the NVIDIA AI Blueprint for Retail Shopping Assistant incorporates NeMo Guardrails microservices to help digital shopping Announced to create more reliable and controlled customer interactions during experiences.
Consulting leaders Taskus, Tech Mahindra, and Wipro are also integrating NeMo Guardrails into their solutions to provide more secure, reliable, and controlled generative AI applications to their enterprise clients.
NeMo Guardrails is open and extensible, providing AI observability and development tools as well as integration with a robust ecosystem of leading AI safety models and guardrail providers. Supports ActiveFence’s integration with ActiveScore to filter harmful or inappropriate content within conversational AI applications, providing visibility, analytics, and monitoring.
Hive provides AI-generated content detection models for image, video, and audio content as NIM microservices that can be easily integrated and tuned into AI applications using NeMo Guardrails.
The Fiddler AI Observability platform easily integrates with NeMo Guardrails to power your AI guardrail monitoring capabilities. Weights & Biases, an end-to-end AI developer platform, is also expanding the capabilities of W&B Weave by adding integration with NeMo Guardrails microservices. This enhancement builds on Weights & Biases’ existing NIM integration portfolio for delivering optimized AI inference in production environments.
NeMo Guardrails offers open source tools for AI safety testing
Developers ready to test the effectiveness of safeguard models and other rail enforcements can use NVIDIA Garak, an open source toolkit for LLM and application vulnerability scanning developed by the NVIDIA Research team. .
Garak allows developers to identify vulnerabilities in systems that use LLM by assessing issues such as data leakage, prompt injection, code hallucinations, and jailbreak scenarios. Garak enables developers to detect and address potential weaknesses in AI models, making them more robust and secure by generating test cases with incorrect or inaccurate output.
availability
NVIDIA NeMo Guardrails microservices, NeMo Guardrails and NVIDIA Garak toolkits for rail orchestration are now available to developers and enterprises. Developers can start building AI safeguards for their customer service AI agents using NeMo Guardrails with this tutorial.
Check out our software product information notice.
