[AI Brief] ITOCHEN Group Launches Enterprise AI Security Solution 'AgentGo Guard'

IT Daily runs an 'AI Brief' section that compiles major AI news from Korea and overseas. ITCEN Group said on the 1st that it had launched the enterprise AI security solution 'AgentGo Guard'.

'AgentGo Guard' is an integrated AI guardrail security solution for existing AI agent and RAG environments. Its key feature is that it can be applied without separate system overhauls.

ITCEN Group said the launch is intended to support enterprises and public institutions in adopting generative AI safely. It also set out support for building an agent ecosystem as another goal.

Earlier, ITCEN CLOIT had pre-released 'AgentGo 2026.' 'AgentGo 2026' is a multi-AI-agent management platform.

The platform comes with features for integrated management and connection of individual AI agents, as well as support for actual work execution. Together with 'AgentGo Guard,' ITCEN Group is moving to build the related framework.

AgentGo Guard is a standalone security platform that can be linked with AgentGo 2026 and is designed to integrate with enterprises' existing AI agent and RAG environments. The company said the security layer can be flexibly applied regardless of each company's AI infrastructure configuration.

The platform has a layered defense system and an adversarial attack defense system. It operates a multi-stage detection system that combines prohibited-word filtering, regular-expression filtering, natural language-based context analysis, and LLM-based guardrails, while identifying and blocking 'echo chamber' attacks that gradually weaken defenses through repeated conversations on a session-by-session real-time basis. It also identifies and blocks sophisticated bypass and jailbreak attempts in real time at the session level. It also includes Korean-language-specific data loss prevention (DLP) features and global-standard data loss prevention (DLP) features.

Before deployment in a live operating environment, the company provides a pre-simulator. This allows security policies and attack scenarios to be pre-validated in a virtual sandbox, while the optimal policy level between false positives and false negatives can also be precisely tuned in advance.

Superb AI said on the 1st that it had developed an AI model for early detection of electric vehicle fires in underground parking lots through a joint study with the National Fire Research Institute.

The model is designed to detect the faint smoke in the early stage of battery thermal runaway, with a focus on spotting electric vehicle fires in underground parking lots at an earlier stage.

As a result, detection time was cut by 65% compared with existing methods, and the false alarm rate was reduced by 80%.

The implementation takes the form of a real-time monitoring system running on an edge device, analyzing on-site video without going through the cloud and making use of existing CCTV infrastructure.

When a fire is detected, it provides on-screen warnings and sound alarms and records detection logs. It can also be accessed from a PC on the same network to the local web server, allowing detection status to be checked immediately without an external server.

Photo credit: Superb AI.

Superb AI conducted a proof-of-concept test in an underground parking lot in November last year. The purpose of this experiment was to verify smoke-detection performance under uncontrolled on-site conditions.

Then in August this year, it completed a proof-of-concept test that reproduced an actual fire scenario. This experiment verified smoke-detection performance.

The final phase of research is now under way. The goal is to make the system lightweight enough to run in real time on a small edge device and to complete a field-ready version.

After the research is completed, the company plans to derive quantitative testing standards. Examples of such standards were given as how many meters away a system can recognize something within how many seconds, and these will be used to establish performance criteria for electric vehicle fire detection systems.

Meanwhile, photo credit: Turing. Turing said it had raised Series B funding worth KRW 7 billion and plans to accelerate the global expansion of GPAI, an AI agent specialized in science and engineering, while broadening the service from science and engineering majors to all university students regardless of field.

Turing announced on the 1st that it had raised KRW 7 billion in Series B funding. New investors in the round included Korea Development Bank, Kingo Investment Partners, and Company K Partners, while existing investor Kolon Investment also participated in follow-on funding.

According to Turing, investors focused on GPAI's growth pace and its potential for global expansion. As generative AI quickly becomes an essential learning tool on college campuses, GPAI is seen as having strong potential for adoption.

Turing used the investment as an opportunity to present a new vision for GPAI. The core of the new vision is to expand from an AI agent specialized in science and engineering majors to an AI service for all university students worldwide, regardless of major.

As a practical step in that direction, Turing made all existing paid features completely free in time for the start of the semester at major universities. Photo credit: Younglimwon Soft Lab.

Younglimwon Soft Lab announced on the 1st that, in cooperation with Shinhan Bank, it has introduced a promotion designed to support digital transformation and accounting efficiency for public-interest foundations while easing the burden of ERP adoption.

The support targets 'SystemEver Nonprofit,' a cloud ERP dedicated to public-interest foundations. Younglimwon Soft Lab will discount the monthly fee for 'SystemEver Nonprofit' by KRW 100,000 per month for new customers and apply it for 6 months.

The total discount benefit is up to KRW 600,000. If the promotion is applied, the maximum monthly fee for 'SystemEver Nonprofit' will fall from KRW 280,000 to KRW 180,000.

There are 3 conditions for eligibility. They are: new signup for 'SystemEver Nonprofit'; opening a new Shinhan Bank account dedicated to public-interest foundations or converting an existing account; and new signup for the 'Shinhan Bank Bankin platform.' The support is available to public-interest foundations that satisfy all 3 conditions.

The benefits are offered on a first-come, first-served basis. Younglimwon Soft Lab said it will carry out discount support together with Shinhan Bank to lower the cost of introducing ERP systems for public-interest foundations.

ElevenLabs said on the 1st that MatrixCloud had adopted its AI agent platform 'ElevenAgent,' adding that MatrixCloud is now building a next-generation AI contact center (AICC) counseling environment in earnest. The deployment is based on ElevenLabs' 'voice AI' technology to support AICC implementation.

MatrixCloud said that after introducing the solution, it overcame the latency limits of its existing voice AI as well as the development complexity limits of its existing voice AI. With this as a springboard, MatrixCloud is also pursuing entry into the global AI contact center market.

ElevenLabs said it provides voice AI technology based on a single conversation flow. It also cited as benefits integrated design at the agent level, ultra-low-latency responses for real-time consultations, and voice quality that is difficult to distinguish from a human voice.

It also cited multilingual voice design for brand voices and flexible API expandability for CRM and CCaaS environments such as Salesforce and Zendesk as additional benefits. Through this, it broadened the scope of implementation and integration across the counseling environment.

MatrixCloud said ElevenLabs made a decisive contribution to improving the completeness of customer conversations. It also said ElevenLabs made a decisive contribution to reducing the burden of building and operating enterprise counseling systems.

Intellivix announced on the 1st that it will supply an AI video analytics system combining vision AI and VLM to Shinsegae Duty Free's Myeongdong store.

The system is used at Shinsegae Duty Free's Myeongdong store to detect escalator reverse movement, strollers entering prohibited areas, and risky behavior related to luggage use. When a dangerous situation is detected, it sends immediate alerts through public-address announcements and warning devices, supporting rapid response by control-room staff and helping prevent accidents.

It also detects sudden customer collapses to support the initial response of operators and security personnel.

It also analyzes visitor density in spaces with heavy customer traffic, such as main entrances and elevator halls. This allows early detection of congestion and can be used for proactive safety management, including staffing adjustments and customer-flow redirection.

The system deployed at Shinsegae Duty Free is structured so that vision AI first detects anomalies and VLM then precisely analyzes the surrounding situation and behavioral context.

The company said this approach reduces false detections and false alarms in existing video surveillance, eases monitoring fatigue for on-site security personnel, and helps secure the golden time in emergencies.

Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=241325

References

• https://www.itdaily.kr/news/articleView.html?idxno=241325