HITCEN Group Launches AI Security Solution 'AgentGo Guard'
TECHWORLD ·
✦ AI Summary
HITCEN Group announced the launch of 'AgentGo Guard (AgentGo Guard),' an integrated AI guardrail security solution.
The solution detects and blocks generative AI security threats such as prompt injection, sensitive information leaks, and jailbreaks, and connects via API to in-house applications, RAG, and AI agent workflows.
It also provides DLP, sandbox testing, call log encryption, and anomaly alert functions, and can be deployed in public cloud and Docker and Kubernetes environments in closed networks.
HITCEN Group announced on the 1st that it had launched 'AgentGo Guard (AgentGo Guard),' an integrated AI guardrail security solution. The launch is aimed at protecting generative AI and AI agent environments for companies and public institutions.
HITCEN Group introduced 'AgentGo Guard' as a product designed to address security threats that companies and public institutions may face as the use of generative AI and AI agents spreads. 'AgentGo Guard' is an enterprise platform that detects and blocks security threats in the process of using generative AI, including prompt injection, sensitive information leaks, and jailbreaks.
The solution was developed as a standalone platform and connects via API. It can be applied separately to existing enterprise AI environments, with integration targets including in-house applications, retrieval-augmented generation (RAG), and AI agent workflows.
HITCEN Group previously introduced a multi-AI agent management platform, 'AgentGo 2026 (AgentGo 2026),' in January through its affiliate HITCEN CLOIT. 'AgentGo Guard' supports integration with 'AgentGo 2026 (AgentGo 2026)' and also supports integration with third-party AI agent platforms.
Security threats are identified not through a single filter but through multiple layers of defense. The security threat detection structure first passes through prohibited-word and regular expression filtering, then applies natural language context analysis and LLM-based guardrails, and afterward determines whether an attack is underway.
It also separately monitors bypass attacks that use accumulated conversations. Session-based detection targets include Echo Chamber attacks, which are attempts to bypass security systems through multiple rounds of conversation, as well as jailbreak attempts, and it tracks attacks based on repeated interactions. The purpose is to control AI agents' operations within their permitted scope.
Monitoring functions for information leak risks have also been added. The included function provides DLP, and DLP supports the Korea Financial Telecommunications & Clearings Institute's CMS account number format and the Ministry of Land, Infrastructure and Transport's vehicle number format. It also detects and anonymizes personal information such as names, addresses, and resident registration numbers.
Before actual deployment, it provides simulation testing of attack scenarios in a separate sandbox before security policies are applied. The sandbox is designed to verify detection policies, and the expected benefits are reducing false positives that block normal requests and reducing false negatives that miss attacks.
At the operational stage, it manages call logs and anomaly signs. Call logs are stored and managed using AES-256 encryption, and administrators must enter a reason when viewing the logs. For anomaly signs, alerts are sent to the person in charge when configured detection thresholds are exceeded.
The deployment environment can be selected to fit enterprise AI infrastructure. Supported environments include public cloud and Docker and Kubernetes in closed networks. Accordingly, it can be applied to companies and institutions operating internal RAG and business AI agents, as well as to finance and public-sector fields with high security requirements.
Jeong Seung-mo, head of the AI Innovation Center at HITCEN Global, said that AgentGo Guard is an enterprise AI defense line that preemptively controls unethical and hateful AI expressions, prompt injection, confidential information leaks, and false information caused by hallucinations. He also said the company intends to support the safe use of AI by businesses.
Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=406349
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.