跳到正文
The Elec· sung ji on·· 3 天前AI 評分36

Coontec 採用 Nvidia DGX B300 進行內部 AI 安全驗證

Coontec Adopts Nvidia DGX B300 for In-House AI Security Validation

AI 導讀

Coontec 採用 Nvidia DGX B300 AI 超級運算系統,提升其 AI 智能體安全平台 ASH(Agentic Security Harness)的威脅偵測模型,並支援 AI 安全研究。

正文

Nvidia’s DGX B300 AI supercomputing system. (Photo: Nvidia)

Nvidia’s DGX B300 AI supercomputing system. (Photo: Nvidia)

Nvidia’s DGX B300 AI supercomputing system. (Photo: Nvidia)
Nvidia’s DGX B300 AI supercomputing system. (Photo: Nvidia)

Coontec said on Nov. 19 it will use Nvidia’s DGX B300 AI supercomputing system to enhance threat detection models for its AI agent security platform, ASH (Agentic Security Harness).

Coontec has recently expanded its software bill of materials (SBOM) security business into AI supply chain security, known as AI-BOM. An AI-BOM is an inventory that tracks the components and origins of AI models and training data.

Coontec adopted the system to meet growing computational demands for AI security research. Running open-weight models with hundreds of billions of parameters and examining their internal operations requires dedicated hardware equipped with large-capacity graphics processing unit (GPU) memory.

The DGX B300 is equipped with eight Nvidia Blackwell Ultra GPUs, each featuring 288 gigabytes (GB) of high-bandwidth memory 3e (HBM3e). Total GPU memory exceeds 2 terabytes (TB). According to Nvidia, the system delivers 72 petaflops (PFLOPS) of AI training performance and 144 PFLOPS of inference performance.

Another advantage is the ability to process AI models, training data and attack scenarios internally. This is particularly relevant to financial institutions and customers in the public and defense sectors that require isolated networks and on-premises environments.

Coontec plans to gradually incorporate results obtained using the DGX B300 into ASH. Dynamic analysis results from large language models (LLMs) will be applied to runtime protection during AI agent execution, while AI-BOM analysis results will be integrated into model supply chain verification. The company subsequently plans to expand its business into validation services that assess AI model risks in advance.

“As AI spreads across industries, AI models and agents themselves are becoming new attack surfaces,” said Bang Hyuk-jun, chief executive officer of Coontec. “With the introduction of the DGX B300, we can now handle the entire process of developing and validating AI using our own capabilities.”

來源:The Elec · thelec.net