Ezserve Launches Endpoint-Based Next-Gen AI Firewall 'EZRO-AI FIREWALL'

Ezserve has officially launched its generative AI security platform, EZRO-AI FIREWALL. This new product integrates security measures to protect internal LLMs from external attackers and control employees' use of external AI services into a single platform. As AI technology expands into planning, development, and HR functions across the organization, security strategies are shifting from mere adoption to ensuring actual control. The AI Basic Law, enacted in January 2024, mandates safety measures for high-impact AI use, and the revised National Cybersecurity Framework (N2SF) allows limited AI adoption in public and financial sectors, making secure operational environments critical. Internal LLMs, connected via Retrieval-Augmented Generation (RAG) to internal databases, have become new hacking targets. Unlike traditional attacks, attackers now steal system prompts or extract reference data through natural language queries, making detection difficult with conventional WAFs. Ezserve's new product uses agents on user endpoints to inspect bidirectional queries and responses to internal models. It blocks prompt injection and malicious command attempts as per OWASP LLM TOP 10, and controls responses containing internal data or personal information. It applies the same checks to external services like ChatGPT and Claude, eliminating blind spots by performing inspections on endpoints, even in remote work or hotspot environments. The product employs a multimodal detection engine to automatically identify and anonymize personal and confidential data in text and images, passing it to AI models without losing context, thus maintaining productivity. It also includes diagnostic tools to visualize AI usage by department and user threat points via dashboards, enabling executives to measure technology investment returns and implement related governance frameworks.