[AI Revolution](27) Security Threats from ChatGPT, We Will Catch Them... Jiranjigyo Data

Jiranjigyo Data Prevents ChatGPT Personal Information Leakage with 'AI Filter'

Recently, more companies are trying to utilize OpenAI's 'ChatGPT' for their work. However, many are reluctant to adopt ChatGPT due to concerns about internal information leaks. Generative AI services based on large language models (LLM) like ChatGPT respond to user queries based on trained data. They continuously improve the quality of their answers by learning from user questions and evaluations of the responses. However, since they learn from user queries, there is a risk that confidential corporate information could be learned and exposed to an unspecified number of people.

In fact, a domestic company experienced an internal information leak caused by the use of ChatGPT. It was revealed during internal monitoring that ChatGPT was used to resolve a program error. The company notified employees to be cautious when using ChatGPT. They also implemented measures to prevent recurrence, such as limiting upload sizes during queries. Additionally, some departments banned the use of ChatGPT.

Data protection specialist company Jiranjigyo Data has launched an AI-based data protection solution called ‘AI Filter (FILTER)’ to prevent leaks of confidential information and personal data caused by misuse of ChatGPT.

AI Filter is a product that integrates ▲technology for detecting and de-identifying personal and confidential information within unstructured text data ▲technology for detecting and de-identifying personal information in image and video data ▲technology for generating augmented training data through synthetic data ▲filtering technology for queries and responses to protect confidential and personal information in AI services such as ChatGPT. It is provided in the form of an API (Application Programming Interface), allowing integration and use within organizational systems.

A key feature of AI Filter is that it is offered as an API. It can be linked with Jiranjigyo Data’s data and personal information protection product line ‘Filter Series’ as well as various solutions such as collaboration tools and groupware to prevent leaks of personal and critical corporate information.

AI Filter System Configuration Diagram. Provided by Jiranjigyo Data

Jowonhee, CEO of Jiranjigyo Data, set two goals to respond to security threats arising from new IT technologies like ChatGPT. One is to develop filtering technology that prevents input and leakage of confidential and personal information, like AI Filter. The other is to advance protection technology that safeguards personal and confidential information during data utilization processes such as AI training.

CEO Jo explained, "We are continuously advancing de-identification technology that supports preventing personal and confidential information from being compromised during AI training processes," adding that ‘ID Filter’ was created to support safe data utilization. Released in 2021, ‘ID Filter’ is a personal information de-identification solution. It converts personal information into pseudonymous or anonymous data that cannot identify individuals, enabling data use without privacy infringement. As interest in data utilization grows day by day, demand and attention for this solution are also rapidly increasing.

The reason Jiranjigyo Data could respond quickly to ChatGPT-originated security threats is due to over 15 years of accumulated expertise in the industry. Since before the enactment of the Personal Information Protection Act, starting in 2006, Jiranjigyo Data has been releasing personal information and data protection solution products to help institutions and companies securely protect data and personal information.

Beginning with the launch of the personal information and inappropriate content filtering solution ‘WebFilter’ in 2006, followed by the personal information protection solution ‘PCFilter’ in 2010, the server personal information diagnosis solution ‘ServerFilter’ in 2011, and the ‘ID Filter’ in 2021, Jiranjigyo Data has been providing the ‘Filter’ series of data and personal information protection solutions.

Jiranjigyo Data plans to continue research and development of AI-based personal information protection technologies. CEO Jo stated, "Personal and confidential information is stored in numerous forms of data. To identify and protect personal and confidential information within vast amounts of data, AI technology is essential," adding, "We will continue AI technology research to safely and conveniently protect personal and confidential information contained not only in databases and document files but also in various data such as images and videos."