Nvidia is launching three new NIM microservices—small, independent services within larger applications—to provide enterprises with enhanced control and safety for their AI agents.
The first microservice focuses on content safety, preventing AI agents from generating harmful or biased outputs. The second ensures conversations remain confined to approved topics, and the third helps safeguard AI agents against jailbreak attempts that could bypass software restrictions.
These new microservices are part of Nvidia NeMo Guardrails, the company’s open-source toolkit designed to help businesses strengthen their AI applications.
The move reflects a growing awareness among AI companies that enterprise adoption of AI agent technology may be slower than initially anticipated. While Salesforce CEO Marc Benioff recently predicted that over a billion agents will run on Salesforce within the next year, the reality is likely to be more measured.
A Deloitte study found that only 25% of enterprises are using or plan to use AI agents by 2025, with the figure rising to about 50% by 2027. This indicates significant interest in AI agents but reveals a slower adoption rate compared to the rapid pace of AI innovation.
By introducing these new tools, Nvidia aims to make AI adoption more secure and practical for enterprises, though it remains to be seen whether this will accelerate the integration of AI agents across industries.
