联想韩国与韩国文档AI初创Lomin合作推出基于ThinkSystem服务器+英伟达GPU的非识别人一体化设备，主攻金融/公共本地部署市场

Lenovo Korea is partnering with document AI company Lomin to introduce the Lomin Redact AI Appliance, a server-based solution designed to de-identify personal information using artificial intelligence.
The appliance integrates AI computing infrastructure, de-identification software and an operating environment built on Lenovo ThinkSystem servers and Nvidia GPUs.
Lomin Redact automatically detects personal and sensitive information and masks it to prevent individuals from being identified. The solution identifies and conceals personally identifiable information (PII) contained in various documents and images, allowing companies to de-identify sensitive data without transferring original documents outside their organizations.
Lenovo Korea said the integrated solution enables companies and institutions handling large volumes of documents to deploy AI infrastructure and specialized software in a single environment while meeting data security and regulatory requirements.
KB Kookmin Card, the General Insurance Association of Korea and Korea Housing & Urban Guarantee Corporation are already using Textscope PrivacyGuard, the predecessor to Lomin Redact.
Lenovo Korea said the partnership expands customer options and focuses on reducing the complexity of deploying AI infrastructure in regulated industries, including finance and the public sector, where data location and access controls are critical.
The company plans to target on-premises and isolated-network environments in industries handling sensitive information, including finance, government and healthcare.
Data de-identification involves removing or transforming identifying information, such as names and telephone numbers, to make individuals difficult to identify.
As AI systems train on and analyze large datasets, personal information may be exposed or combined with other data to re-identify individuals. De-identification technology can help organizations increase the use of data for AI while protecting personal information and meeting regulatory requirements.
Lomin Redact uses a vision-language model (VLM) to analyze document layouts, text and context, enabling it to detect both structured patterns and unstructured personal and sensitive information.
The solution is designed to process various types of unstructured documents, including handwritten text, information within tables, areas overlapping with stamps, and low-quality or rotated documents.
It also calculates confidence scores for de-identification results, identifies areas requiring human review and supports role-based access controls to manage de-identification targets and review scope.
Lenovo Korea plans to expand enterprise AI applications by combining its AI infrastructure with specialized AI software developed by South Korean companies.
The company said it would strengthen collaboration with domestic AI partners to help customers deploy AI in real-world business environments, particularly in industries where data security and control are essential.