Korea’s AI Ecosystem Advances and Secures Agentic AI With NVIDIA

As AI agents become more capable, they require advanced open models and robust security measures to operate safely. Hyundai Motor Group is investigating NVIDIA's NemoClaw frameworks and the NVIDIA OpenShell runtime environment for this purpose. Meanwhile, AIM Intelligence is incorporating its Starfort platform with NVIDIA OpenShell to provide secure AI agents for increasing employee productivity and streamlining enterprise workflows.

OpenShell is a crucial part of NVIDIA's Open Agent Safety Platform, delivering a secure runtime environment equipped with sandboxed execution and policy-based controls to protect sensitive data, regulate access to enterprise tools, and govern the behaviors of autonomous agents.

AIM Intelligence, a prominent Korean AI security firm, is collaborating with NVIDIA OpenShell by integrating its Starfort agent control platform. This partnership aims to merge OpenShell's runtime access controls with Starfort's ability to detect malicious instructions and prevent data leakage. The company is also contemplating a pilot project targeting the public energy sector, with potential applications extending to enterprise, industrial, and robotics sectors.

On the model side, LG AI Research is making its K-EXAONE-2.5 application programming interface accessible, created using NVIDIA Nemotron datasets. This dataset includes Nemotron-Personas-Korea, an open collection of 7 million synthetic Korean personas. These personas are grounded in official Korean statistics and serve as a foundation for model builders to ground their training data and evaluation processes within the Korean context.

NVIDIA Nemotron offers a suite of open models, datasets, and techniques designed to assist developers in constructing specialized, efficient AI agents. For more information on NVIDIA Nemotron open models and guidance on building secure agents with NVIDIA OpenShell, visit NVIDIA's official resources.