Security

Kuntech Introduces NVIDIA's DGX B300, Builds AI Security Research Infrastructure

IT DAILY ·

Kuntech Introduces NVIDIA's DGX B300, Builds AI Security Research Infrastructure
NVIDIA's DGX B300 [Photo: NVIDIA]

✦ AI Summary

Kuntech said on the 19th that it has built an AI security research infrastructure by introducing NVIDIA's DGX B300 AI supercomputing system.

The infrastructure is intended to support in-house AI model reinforcement learning, LLM security analysis, and AI supply chain verification without relying on external cloud services.

Kuntech plans to use the infrastructure for reinforcement learning, parallel execution of open-weight models, LLM dynamic analysis, and AI-BOM-based supply chain verification in ASH.

BizCrushWant to follow Korean talks and live streams in real time?Try CapCrush on the web

Kuntech has been doing business in security areas including software supply chains, infrastructure, and OT. In 2024, it developed its own SBOM engine, and more recently it unveiled an AI agent security platform, ASH.

Kuntech announced on the 19th that it has built an AI security research infrastructure by introducing NVIDIA's DGX B300 AI supercomputing system. With this introduction, the company can now carry out AI model reinforcement learning, LLM security analysis, and AI supply chain verification in-house without using external cloud services.

Behind the decision to build this in-house infrastructure were the growing compute demands of AI security research and the limitations of operating on an external cloud basis. The company also needed dedicated equipment equipped with large-capacity GPU memory for running open-weight models and observing their behavior.

In addition, financial, public-sector, and defense customers have required closed-network and on-premises environments. Kuntech's decision to build its own infrastructure is a response to cloud limitations, the need for dedicated equipment, and customer environment requirements.

DGX B300 is a system equipped with eight NVIDIA Blackwell Ultra GPUs. It has 288 GB of HBM3e per device, and total memory exceeds 2 TB. NVIDIA describes the system's AI training performance as 72 PFLOPS and its inference performance as 144 PFLOPS.

Kuntech outlined three areas for using the AI infrastructure it built. The first use is repeated reinforcement learning training for the threat detection model embedded in ASH. By using analysis results as reward signals in internal training, the company aims to improve detection accuracy and inference quality.

The second use is parallel execution of large open-weight models, and the third is LLM dynamic analysis. These tasks are designed to check for hidden risks before enterprises adopt public models and before they deploy AI agents, and they involve running multiple models at the same time while observing internal operations and outputs. In this process, the company verifies responses to attacks such as prompt injection and jailbreaking.

Kuntech uses ASH's Model Gate feature to perform provenance tracking analysis based on AI-BOM. This makes it possible to trace the source and composition of training data before a model is brought in, and to check for data contamination and backdoors.

Kuntech plans to apply the research results to ASH in stages. It will feed the results of LLM dynamic analysis into runtime protection features that block attacks during agent execution, and it will apply AI-BOM-based analysis to Model Gate's supply chain verification functions. It then plans to expand the business into a service that preemptively diagnoses and verifies the risk level of AI models adopted by companies and institutions.

Kuntech CEO Bang Hyeok-jun said that, as AI spreads, models and agents are becoming new attack surfaces. He added that this equipment introduction has enabled the company to carry out the entire AI verification process with its own capabilities. He said the goal is to establish itself as a leading company in AI supply chain security based on this foundation.

Source: IT DAILY · Kim Ho-jun
Original: https://www.itdaily.kr/news/articleView.html?idxno=243061

References

This article was produced with the help of an automated content generation algorithm.


Source: IT DAILY

View original

This article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.

Keep reading

BizCrush
BizCrush showing live captions over an online meeting

BizCrush app

Live captions and summaries for meetings and calls

BizCrush Inc. · CEO Taemin Kwak

HQ 100 Middlefield Rd, Menlo Park, CA 94025, USA

Korea office Z112, Room 602, 6F, 602 Yeongdong-daero, Gangnam-gu, Seoul, Republic of Korea

Support help@bizcrush.ai · Partnerships business@bizcrush.ai

AI Notice

Some features use generative AI.

Those features carry their own pre-use notice and mark AI-generated results. AI output is for reference only — always check the original source and consult an expert before making an important decision.