AI changed the security perimeter. Nati Hazut, CEO and Co-Founder of Bold Security, says work has moved from browsers and SaaS applications to interactions that happen on devices. That shift has broken assumptions built into many security products.
Bold Security develops endpoint data protection technology built around small language models running directly on devices. Its platform lets security teams coach users, redirect activity, or block risky actions across browsers, desktop applications, and AI agents.
That product focus follows a basic problem: the endpoint is now where data interactions meet. Network inspection, cloud application controls, and storage monitoring each see a slice of activity, but the device connects those slices in one place.
AI Changed Where Work Happens
“The main thing that AI has done is move where work happens,” Hazut said. Sensitive activity once handled through browsers and SaaS applications now also happens through on-device interactions. That change matters because desktop applications can read files directly off the device and encrypt their traffic.
Network inspection cannot see traffic that never routes through the network path it monitors. As Hazut put it, “Network inspection only sees traffic that routes through it, and desktop applications encrypt theirs and bypass the proxy.” A proxy cannot inspect what the application avoids sending through it — an awkward limitation for products designed around traffic visibility.
Cloud security tools face their own boundaries. CASB sees sanctioned cloud apps, while DSPM sees data sitting in cloud storage. Neither view covers every interaction that takes place between a user, a local file, a desktop application, and an AI tool.
Bold Security places endpoint data protection at that intersection. Its technology uses small language models on the device, allowing security teams to respond to activity across browsers, desktop applications, and AI agents rather than focusing on one application category.
Shadow AI Is Now a Data Control Problem
Shadow AI includes more than employees opening personal ChatGPT accounts. It also covers desktop applications, browser extensions, APIs, Model Context Protocol connections, and autonomous agents.
A year ago, the concern was shadow AI itself: employees were using AI tools outside approved systems. The focus has now moved to controlling the data going through those tools, because the risk involves files and information moving through channels that security teams may not monitor.
Desktop applications make that control difficult. They can read files directly off the device and encrypt their traffic, leaving network inspection with no reliable view of what the application accessed or where the data went.
That is why the endpoint matters in Hazut’s view. It is the one place where all data interactions come together, including browser activity, desktop applications, AI agents, local files, and connections to external tools.
The result is a security problem that cannot be solved by adding another view of sanctioned cloud software or stored cloud data. Security teams need controls that operate where the interaction happens — on the device — and can coach users, redirect activity, or block risky actions.
Hazut previously co-founded SAM Seamless Network and Polyrize. At Bold Security, he serves as CEO and Co-Founder alongside co-founders Hadar Krasner and Omri Mallis, with the company focused on protecting data at the endpoint as AI changes how work gets done.
The premise is not that browsers, SaaS applications, networks, or cloud storage no longer matter. It is that they no longer tell the whole story. AI has widened the route data can take, and the endpoint has become the place where those routes converge.
