Infoseeds, an ontology-based enterprise AI company, said on Tuesday it launched a graph-based AI security solution called GOS (Graph Optimized Security) and supplied it to a major South Korean semiconductor company.

The company said GOS analyzes large-scale security logs using Infoseeds' in-house Graph ML technology. It then sends only candidates with high threat potential to an Analyzer LLM to verify threat likelihood and behavioral context.

Unlike traditional security solutions that look for threats based on pre-registered attack types and detection rules, GOS learns an organisation's work patterns. It maps relationships among data into graphs and identifies security threats when it detects behaviour that differs from usual patterns.

GOS can be used alongside a company's existing security information and event management (SIEM) and security orchestration, automation and response (SOAR) solutions without replacing them. The company said analysing all logs with commercial large language models (LLMs) could incur costs of trillions of won a year, and it stressed that GOS narrows LLM analysis targets to about 0.068 percent of the full behaviour graph, sharply cutting operating costs while maintaining detection accuracy.

Infoseeds operates its business around Axiom, an ontology-based enterprise AI solution, and DELTAflow, which reads industrial drawings and converts them into knowledge graphs.

Infoseeds CEO Jeong-su Tak (탁정수) said the core of corporate security is not generating more alerts but building an accurate decision framework that does not miss real threats while not disrupting normal work. He said GOS has been recognised for technical accuracy and commercial viability in real enterprise environments, and the company plans to expand deployments to enterprise companies with large-scale security logs and complex work environments, starting with this project award.