NVIDIA Details Korea’s Push for Secure Agentic AI
Korean companies are developing secure AI agents and locally grounded models with NVIDIA technologies, including the NVIDIA OpenShell runtime, NemoClaw blueprints and Nemotron datasets. Hyundai Motor Group, AIM Intelligence and LG AI Research are among the organizations involved.
- Hyundai Motor Group is exploring NVIDIA NemoClaw blueprints and OpenShell.
- AIM Intelligence is integrating Starfort with NVIDIA OpenShell.
- OpenShell uses sandboxed execution and policy-based controls.
- LG AI Research is opening API access for K-EXAONE-2.5.
- Nemotron-Personas-Korea contains 7 million synthetic Korean personas.
Security controls for autonomous agents
As AI agents take on more work, NVIDIA says they require both capable open models and security that develops alongside them. Hyundai Motor Group is exploring NVIDIA NemoClaw blueprints and the NVIDIA OpenShell runtime as it evaluates approaches for agentic AI.
OpenShell is a key component of the NVIDIA Open Agent Safety Platform. The runtime uses sandboxed execution and policy-based controls to protect sensitive data, manage access to enterprise tools and govern the actions of autonomous agents. These functions are designed to address the risks that arise when agents can perform tasks across business systems.
AIM Intelligence combines runtime controls with threat detection
AIM Intelligence, a Korean AI security company, is integrating its Starfort agent control platform with NVIDIA OpenShell. The companies’ approach is intended to combine OpenShell’s runtime access controls with Starfort’s ability to detect malicious instructions and data leakage.
AIM Intelligence is also considering a preliminary proof of concept for the public energy sector. According to NVIDIA, the technology could have applications across enterprise, industrial and robotics environments. The announcement does not provide details about the timing or scope of the potential proof of concept.
LG AI Research opens access to K-EXAONE-2.5
On the model side, LG AI Research is opening application programming interface access for K-EXAONE-2.5. The model was developed using NVIDIA Nemotron datasets, including Nemotron-Personas-Korea.
Nemotron-Personas-Korea is an open dataset containing 7 million fully synthetic Korean personas. NVIDIA says the personas are grounded in official Korean statistics, allowing model builders to ground training data and evaluation in Korean contexts.
NVIDIA Nemotron is a family of open models, datasets and techniques intended to help developers build specialized, efficient AI agents. For users and organizations in Korea, the announcement points to a combination of locally relevant model resources and controls for managing agent behavior. NVIDIA’s examples also show the technologies being considered across automotive, enterprise, energy, industrial and robotics settings.
