Nvidia has introduced three new NeMo Inference Microservices (NIM), significantly accelerating industry use of AI agents.
These microservices, part of the Nvidia NeMo Guardrails ecosystem, solve essential concerns about deploying AI agents, including safety, security, and compliance.
The first Microsoft microservice focuses on content safety, ensuring that AI-generated output is free of dangerous or biased information. The second focuses on topic control, ensuring that conversations stay within boundaries and do not deviate into unauthorized areas. The third microservice strengthens AI systems against jailbreak efforts, preventing them from being persuaded to circumvent constraints.
These microservices supplement Nvidia's existing NeMo Guardrails, an open-source toolkit aimed at improving the trustworthiness of AI systems. By using lightweight, specialized models, developers may be able to address gaps left by general security measures, ensuring a higher level of precision and control in complex AI workflows.
Adopting these techniques transforms industries such as healthcare, banking, automotive, and retail. For example, Lowe's uses NeMo Guardrails to increase customer interactions by providing the shop personnel with precise product information while assuring conversational safety. Similarly, Cerence AI combines similar microservices to provide safe, context-aware in-car assistance.
Nvidia's approach is consistent with organizations' growing interest in AI agents, but adoption remains hesitant. According to Deloitte, only 25% of enterprises are employing or planning to use AI agents by 2025, with forecasts of 50% by 2027. Nvidia hopes to fill this gap by offering solutions that make AI deployment more trustworthy and flexible.
Nvidia has made the NeMo Guardrails platform open-source to assist developers. This allows it to interact with other AI safety solutions, such as ActiveFence and Fiddler AI Observability. Furthermore, the Nvidia Garak toolset allows developers to test AI systems for vulnerabilities, including prompt injections and data leaks.
By addressing important enterprise concerns, Nvidia positions itself as a leader in developing future-proof, secure, and scalable AI solutions.