BITDEFENDER GravityZone AI Visibility and Control (AIVC) is described as a discovery and risk layer that answers key questions about AI on protected Windows endpoints: which AI software is present, who is running it, and how risky each tool is.
The article outlines four exposure classes that sit outside existing controls: AI services reached via the browser, AI inside development environments (such as coding assistants inside editors and credentials in working directories), AI running locally on the endpoint (local model runtimes, MCP or agent frameworks), and AI with reach into other systems (connections to databases, repositories, or external model services).
GravityZone’s BEST agent collects endpoint metadata during risk scans and does not inspect prompts, chat histories, model weights, credentials, or full file content, aiming to surface configuration risk rather than user data.
AIVC presents two views in GravityZone Console: AI Services, which inventories each discovered tool with a risk score and functional category, and AI Findings, the prioritized view listing endpoint-specific findings with risk scores, affected resources, and mitigation links. The article emphasises practical use: it correlates findings with the relevant tool, enabling SOC analysts to understand why something is risky and where it sits in the environment, rather than obscuring it behind high-level metrics.
Enforcement remains with existing Bitdefender controls, such as Web Content Filtering to block browser-based AI services, custom detection rules to terminate certain tools across endpoints, and PHASR-based hardening to constrain how agents operate. AIVC can be enabled by editing an existing policy; results appear after the next risk scan.