AI agents now run directly on employee and developer machines - Claude Code, Claude CoWork, Codex, OpenClaw and others - with the user’s credentials, the ability to execute commands, and a growing stack of MCP servers and Skills installed alongside them, often with no one approving any of it.
James Berthoty, founder of Latio, and Rani Osnat of Backslash Security look at what this means for security teams: what Latio’s 2026 research and Backslash’s aggregated customer data show about how widespread it already is, where endpoint and network tools lose sight of it, and what a vendor-agnostic control layer has to do to keep agents useful without letting them act unchecked.
Privacy is important to us, so you have the option of disabling certain types of storage that may not be necessary for the basic functioning of the website. Blocking categories may impact your experience on the website. More information
