Agents Gone Local: The Endpoint Risk Security
Teams Can’t Afford to Ignore

Agents like Claude Code, Claude CoWork, Codex, and OpenClaw now run autonomously on employee and developer machines. They inherit local credentials, execute commands, and install MCP servers and skills, often with no human in the loop.

Latio’s 2026 AI Security Market Report finds that 45% of security leaders now rank endpoint agents as their top AI security concern, and the share of teams with dedicated AI security budget has jumped from 8% to 37%.

Backslash’s own aggregated customer data shows why. 79% of organizations reported little to no visibility into their agentic AI inventory, 60% of environments had credential-related risks, and in some cases a single laptop presents multiple policy violations Personal accounts were used to access coding agents in every enterprise environment assessed.

Join James Berthoty, founder of Latio, and Rani Osnat, VP Marketing at Backslash Security, for this live webinar, where they unpack the rapid evolution of agentic AI use and how security must adapt.

‍You’ll learn:

• Endpoint agent blast radius: a “real and present danger”?
• What Latio’s research and Backslash’s aggregated customer data reveal
• Where legacy endpoint protection and network-based tools leave blind spots
• What an effective, vendor-agnostic endpoint control layer looks like