Bold Security Targets Endpoint Blind Spots in AI Workflows

Traditional data loss prevention tools often fail to capture the full scope of AI activity. Desktop applications frequently bypass network proxies, leaving security teams without visibility into how information is processed by agents or command-line interfaces. By shifting the monitoring layer directly onto the device, Bold aims to bridge this gap, analyzing prompts, clipboard activity, and file access in real time without requiring data to leave the user’s machine.

Nati Hazut, co-founder and CEO of Bold, noted that endpoint exposure was once an accepted compromise in cybersecurity, but the rise of AI has transformed these minor gaps into critical business risks. The platform distinguishes between routine operations and high-stakes data exfiltration by evaluating the context of each interaction—identifying not just that an AI tool was used, but exactly what information was involved and where it was directed.

This new functionality covers three primary vectors: web-based copilots, desktop AI applications, and autonomous agents utilizing Model Context Protocol (MCP) payloads. The technology is currently available in private preview, with live demonstrations scheduled for next week at the Black Hat conference in Las Vegas.