Graffer Establishes 'Local LLM Utilization Technology' to Support Document Processing Without Sending Personal Data Off-Premise
Graffer has established a 'Local LLM utilization technology' that processes documents within a user's PC or internal server, without sending personal or confidential information to external servers via the internet. This technology can run on standard PC environments without requiring high-performance dedicated servers. It supports extraction and verification for applications and forms, enabling safe AI usage in government and financial sectors.
📋 Article Processing Timeline
- 📰 Published: June 3, 2026 at 10:00
- 🔍 Collected: June 3, 2026 at 10:25 (25 min after Published)
- 🤖 AI Analyzed: June 3, 2026 at 11:05 (40 min after Collected)
Graffer Inc. (Headquarters: Shibuya, Tokyo; Representative: Daichi Ishii, hereafter 'Graffer') has established a 'Local LLM utilization technology' that allows AI to process documents on a user's PC or internal server without transmitting personal or confidential information to other companies' servers over the internet.
This technology supports the extraction of required items and content verification for images such as application forms, identity documents, bank information, various forms, and surveys. Graffer is advancing technical verification for document processing applications using a local LLM configuration that operates on standard PC environments without requiring high-performance dedicated servers, and has confirmed the technical feasibility of item extraction and verification support.
While configurations using high-performance servers or large-scale models may offer higher accuracy and processing speeds, closing the processing environment locally is a critical requirement for sectors that cannot export personal or confidential information. Graffer will continue to work on improving accuracy, speed, and usability to support document processing that is practical even in configurations that do not require high-performance dedicated servers.
* LLM (Large Language Model): The foundational technology used in AI like ChatGPT for understanding and generating text and images. Common cloud-based generative AI services send input data to external servers for processing. In contrast, 'Local LLM' as referred to in this release means a configuration where processing is completed on the user's PC or within an organization's server, without transmitting data to external providers' servers. This technology utilizes publicly available open models optimized for local environment operation.
Background of Development
Cloud-based generative AI, such as ChatGPT, operates by sending user-input information over the internet to the service provider's servers for processing (hereafter referred to as 'Cloud LLMs'). Consequently, in fields like government, finance, healthcare, BPO, legal, and HR, many cases have had to forgo generative AI adoption because of the rule that 'personal or confidential information cannot be sent to servers outside the organization's control.'
For example, at municipal windows, staff manually verify and input information such as names, addresses, and account numbers from application forms one by one. While operational efficiency through generative AI is anticipated, adoption has often been shelved due to the constraint of not being able to transmit personal information to external servers. In private enterprises, there are numerous operations, such as managing customer-related documents, where data cannot be allowed to leave the company.
Furthermore, since Cloud LLMs often involve usage-based fees, there has been the challenge of unpredictable costs for operations that process large volumes of documents continuously.
Based on this, Graffer established document processing technology utilizing Local LLMs to enable AI usage in environments where data cannot leave the company. This technology uses a configuration that operates on standard PCs without requiring high-performance dedicated servers, and Graffer has confirmed its technical feasibility for extraction and verification. Moving forward, the company will proceed with accuracy, speed, and operational improvements through verification in actual business settings.
Features of the Technology
1. Configuration that completes processing without sending personal/confidential info off-premise
By completing AI processing in the local environment of the user's PC, document images and extracted data can be processed without being sent over the internet to other companies' servers. This configuration is designed with operation in closed environments without internet access or internal company servers (on-premises) in mind, expanding the possibilities for AI utilization based on personal information protection and security policies in sectors where Cloud LLM adoption has been difficult, such as municipal application processing, window services, and BPO operations.
2. Operation without AI usage-based fees or large-scale infrastructure investment
While Cloud LLMs may involve usage-based fees depending on the service, utilizing Local LLMs allows for operation without relying on such fees. This is expected to provide greater cost predictability for large-scale document processing and routine verification tasks.
Moreover, while technology for running AI locally previously required high-performance servers or specialized equipment, this technology validates application to document processing using a configuration capable of running on standard PC environments. This structure, which does not require large-scale infrastructure investment, facilitates test deployments in small departments or at a branch level.
This technology supports the extraction of required items and content verification for images such as application forms, identity documents, bank information, various forms, and surveys. Graffer is advancing technical verification for document processing applications using a local LLM configuration that operates on standard PC environments without requiring high-performance dedicated servers, and has confirmed the technical feasibility of item extraction and verification support.
While configurations using high-performance servers or large-scale models may offer higher accuracy and processing speeds, closing the processing environment locally is a critical requirement for sectors that cannot export personal or confidential information. Graffer will continue to work on improving accuracy, speed, and usability to support document processing that is practical even in configurations that do not require high-performance dedicated servers.
* LLM (Large Language Model): The foundational technology used in AI like ChatGPT for understanding and generating text and images. Common cloud-based generative AI services send input data to external servers for processing. In contrast, 'Local LLM' as referred to in this release means a configuration where processing is completed on the user's PC or within an organization's server, without transmitting data to external providers' servers. This technology utilizes publicly available open models optimized for local environment operation.
Background of Development
Cloud-based generative AI, such as ChatGPT, operates by sending user-input information over the internet to the service provider's servers for processing (hereafter referred to as 'Cloud LLMs'). Consequently, in fields like government, finance, healthcare, BPO, legal, and HR, many cases have had to forgo generative AI adoption because of the rule that 'personal or confidential information cannot be sent to servers outside the organization's control.'
For example, at municipal windows, staff manually verify and input information such as names, addresses, and account numbers from application forms one by one. While operational efficiency through generative AI is anticipated, adoption has often been shelved due to the constraint of not being able to transmit personal information to external servers. In private enterprises, there are numerous operations, such as managing customer-related documents, where data cannot be allowed to leave the company.
Furthermore, since Cloud LLMs often involve usage-based fees, there has been the challenge of unpredictable costs for operations that process large volumes of documents continuously.
Based on this, Graffer established document processing technology utilizing Local LLMs to enable AI usage in environments where data cannot leave the company. This technology uses a configuration that operates on standard PCs without requiring high-performance dedicated servers, and Graffer has confirmed its technical feasibility for extraction and verification. Moving forward, the company will proceed with accuracy, speed, and operational improvements through verification in actual business settings.
Features of the Technology
1. Configuration that completes processing without sending personal/confidential info off-premise
By completing AI processing in the local environment of the user's PC, document images and extracted data can be processed without being sent over the internet to other companies' servers. This configuration is designed with operation in closed environments without internet access or internal company servers (on-premises) in mind, expanding the possibilities for AI utilization based on personal information protection and security policies in sectors where Cloud LLM adoption has been difficult, such as municipal application processing, window services, and BPO operations.
2. Operation without AI usage-based fees or large-scale infrastructure investment
While Cloud LLMs may involve usage-based fees depending on the service, utilizing Local LLMs allows for operation without relying on such fees. This is expected to provide greater cost predictability for large-scale document processing and routine verification tasks.
Moreover, while technology for running AI locally previously required high-performance servers or specialized equipment, this technology validates application to document processing using a configuration capable of running on standard PC environments. This structure, which does not require large-scale infrastructure investment, facilitates test deployments in small departments or at a branch level.
FAQ
What is the local LLM utilization technology established by Grapher?
It is a technology that completes document processing within the user's PC or internal server (on-premises environment) without sending personal information or confidential information to external servers via the internet.
What types of documents does this technology support?
It supports images of application forms, identification documents, account information, various reports, surveys, and more, assisting in extracting necessary fields and verifying content.
Is a high-performance dedicated server required?
No, we are advancing technical validation with a configuration that can operate on a general PC environment with a certain level of computational resources, without assuming a high-performance dedicated server.
Why is a local LLM necessary?
In fields such as administration, finance, healthcare, and human resources, there are security constraints that prevent confidential information from being sent to servers outside the organization, making it difficult to use cloud-based generative AI.
What is the cost difference between local LLM and cloud-based generative AI?
There are no usage-based charges that increase with the number of processed items as seen with cloud LLMs, and no large-scale infrastructure investment is required, making it easier to predict costs for routine operations.