#Japan #Generative AI #Tokyo #HPE #SCUTI

Securing Sensitive Information Using On-Premises Generative AI

In an age where data breaches have become increasingly common, protecting sensitive information is non-negotiable for modern businesses. Scuti Corporation has partnered with HPE to demonstrate the effectiveness of generative AI (GAI) systems in environments completely disconnected from external networks. Their published white paper outlines the verification process and results achieved using HPE's state-of-the-art server, the ProLiant DL380 Gen11.

The Necessity of AI in Business Operations

As companies rush to adopt generative AI to enhance productivity, the challenge of safeguarding confidential information continues to loom large. Many organizations have had to restrict the use of cloud-based AI due to risks surrounding the leakage of sensitive data. This white paper provides insight into how businesses can leverage AI while maintaining stringent security, particularly in departments like management, HR, and digital transformation initiatives.

Background of the Verification: Resolving the Trade-Off Between Security and Usability

To truly harness the power of generative AI, organizations need to employ Retrieval-Augmented Generation (RAG), which integrates internal company knowledge—like manuals, regulations, and past minutes—into the AI’s capabilities. However, uploading this critical information to the cloud presents significant psychological and regulatory barriers. To tackle this issue, Scuti utilized HPE’s powerful hardware infrastructure to verify a completely offline version of secure generative AI.

Robust Infrastructure: HPE ProLiant DL380 Gen11

For their verification, Scuti chose the HPE ProLiant DL380 Gen11, known for its exceptional processing power and reliability. The verification tested performance based on the needs of the RAG system, particularly for local execution of large language models (LLMs) and vector searches.

Key Features:

• - High Computational Power: The system was able to handle demanding tasks while maintaining high performance.
• - Strong Security: Utilizing HPE products implemented with features such as the Silicon Root of Trust ensured comprehensive safety across application and infrastructure layers.

The results indicated that operations in an isolated environment could achieve response speeds comparable to cloud AI, enabling accurate response generation based on organizational knowledge.

Contents of the Technical White Paper

The white paper authored by Scuti delves into detailed evaluations regarding:
1. On-Premises GAI Architecture: How to structure a system in a network-isolated environment.
2. Precision Assessment: A comparison between chat functionalities alone and those augmented with RAG.
3. Analysis by Document Format: Insights about response accuracy trends across different materials, including both Japanese and English text and presentation formats.
4. Guidelines for Practical Implementation: Important considerations for management and HR departments for effective adoption.

Downloading the White Paper

For those interested in exploring the findings in detail, Scuti offers a free download of the white paper. Access it through the following URL: Download White Paper

Future Prospects

Scuti aims to expand upon the insights gained from this verification study to enhance their consulting and implementation services for on-premises RAG systems specifically tailored for clients in financial, manufacturing, and public sectors. This initiative aligns perfectly with their goal to resolve language barriers and knowledge monopolization challenges, thereby increasing competitive edges in digital transformation.

Additionally, Scuti will be showcasing their secure generative AI offerings at the 'AI Expo Spring 2026,' taking place at the Tokyo International Forum from April 7-8, 2026. Attendees are encouraged to visit their booth for more insights.

About Scuti Corporation

Founded in November 2015, Scuti Corporation specializes in providing services to enhance operational efficiency through generative AI technologies, notably RAG and AI-OCR. They provide comprehensive support from consulting to implementation, ensuring businesses leverage cutting-edge AI technology effectively.

Company Information:

• - Name: Scuti Corporation
• - CEO: Tomohide Kakutani
• - Tokyo Office: 5F Daiya Gate Ikebukuro, 1-16-15 Minamiikebukuro, Toshima-ku
• - Headquarters: 2117-3 Yamata, Okayama City, Okayama Prefecture

Established: November 2015
Website: Scuti Official Site