Lingji Technology launched AnyInsight Edge, extending its AI security platform to on-premises environments and AI servers. The product addresses challenges where private LLMs ensure data sovereignty but lack AI Agent capabilities, while cloud AI faces regulatory constraints. AnyInsight Edge enables unified policy management across deployments, integrating private LLMs with cloud models while maintaining inspection, access control, and audit trails.
Enterprises deploying AI face a trade-off: private LLMs keep data internal but limit AI Agent and automation capabilities, while cloud AI provides advanced models and Agents but cannot process data subject to localization or regulatory requirements. Many organizations investing in AI servers and private LLMs find models isolated in internal chat interfaces, unable to connect to ERP, MES, or CRM systems for Agent deployment. Self-integration creates risks as Agents move data across systems without network firewall oversight, forming "Shadow Agents" that bypass access controls. Cloud AI services, while individually secure, fragment policy management across platforms when enterprises use multiple providers, creating "Shadow AI" when employees turn to personal accounts for unapproved tools.
AnyInsight Edge Integrates Private LLMs with Cloud AI Models
AnyInsight Edge delivers a single control plane for centralized policy management executed consistently across cloud and on-premises environments. The platform enables enterprises to operate private models (including Gemma, TAIDE, Mistral) with access to AnyInsight's 350+ ready-to-use AI Agents, automation workflows, and 500+ application connectors, routing to 20+ mainstream cloud models when policies permit. AI servers running AnyInsight Edge function as governed infrastructure with AI Gateway and zero-trust generative AI firewall capabilities, subjecting every request — conversations, file transfers, application integrations — to policy engine inspection and logging. Core data, including company databases, AI conversations, transmission data, access settings, and audit logs, remain on-premises with tiered storage for long-term cold data retention.
The platform applies consistent access controls and detection rules across office locations, remote employees, and multiple on-premises sites, aggregating events and audit records into a single management interface. Cross-site policy synchronization maintains uniform governance without requiring HTTPS decryption at network firewalls. Traditional network firewalls manage traffic and access while zero-trust generative AI firewalls inspect AI interaction content, creating complementary defense layers. Network firewalls block unapproved AI services and route AI traffic to AnyInsight, synchronizing network-layer and content-layer policies with unified event monitoring.
Container-Based Deployment Supports Air-Gapped Environments
AnyInsight Edge deploys as a container independent of hardware design, operating in enterprise data centers or integrated with AI server operating systems. Edge deployment requires no GPU resources, preserving full compute capacity for model inference. Policy management initiates directly from Edge in on-premises environments or synchronizes with cloud platforms. Edge functions as a Policy Enforcement Point (PEP) on the path of on-premises AI traffic, performing real-time inspection and execution. The architecture extends to completely air-gapped network environments, independently managing AI server governance and firewall functions without cloud connectivity.
Agentic Security Model Monitors AI Agent Actions
When AI Agents execute autonomous multi-step tasks, protection scope extends beyond user input and model responses to information accessed by Agents, application tools invoked, and data moved between systems. Since Agent next steps cannot be predetermined, controls must apply progressively during execution — the core principle of Agentic Security. Traffic between Agents and internal systems within data centers bypasses network firewalls, requiring Edge deployment for on-premises enforcement.
AnyInsight employs context awareness and semantic analysis to identify sensitive content including personal data, financial and health information, and cloud credentials, triggering real-time alerts, blocking, or masking. Before each Agent tool invocation via Model Context Protocol (MCP), the system checks policies and authorizes based on role and least-privilege principles. High-risk actions within Agents and automation workflows require human approval. Every event generates log entries exportable as audit evidence.
Application Scenarios Span Manufacturing and Finance Sectors
In manufacturing and semiconductor supply chains, process parameters, yield data, and customer drawings remain on-site while on-premises models use Agents and automation workflows to compile production reports and anomaly analysis, with consistent security baselines applied across multiple facilities. Financial and insurance institutions mask customer personal data before model input, maintaining auditable records of every AI use and Agent action to satisfy regulatory and internal control requirements. Government agencies and medical institutions process official documents and medical records with on-premises models, using the same platform to provide AI interaction disclosure, human review, and complete audit trails, aligning with Taiwan's Artificial Intelligence Basic Act direction and corresponding to EU AI Act requirements for transparency and human oversight.
Lingji Technology CEO Yan Ziqiang stated that twenty years ago, enterprise security aligned with network proliferation when firewalls, intrusion prevention, and VPN integrated into UTM. AI reaches a similar inflection point: models will continuously change, Agents will proliferate, and enterprises need control points that execute policies where AI work occurs. AnyInsight Edge transforms every AI server into such a control point.
FAQ
What does AnyInsight Edge enable for enterprises using private LLMs?
AnyInsight Edge enables enterprises operating private LLMs to access 350+ ready-to-use AI Agents, automation workflows, and 500+ application connectors while maintaining unified policy management across on-premises and cloud deployments. The platform integrates private models with 20+ mainstream cloud models, routing requests according to policy while keeping core data, AI conversations, and audit logs on-premises.
How does Agentic Security differ from traditional AI security approaches?
Agentic Security extends protection beyond user input and model responses to monitor information accessed by AI Agents, application tools invoked, and data moved between systems. Since Agent actions cannot be predetermined, AnyInsight checks policies before each tool invocation via Model Context Protocol, applies role-based authorization with least-privilege principles, and requires human approval for high-risk actions while generating auditable log entries for every event.
