The next frontier in endpoint security: Securing local AI agents with Microsoft Defender (opens in new tab)

AI agents are now doing real work on the endpoint — reading files, running commands, browsing the web, and acting on behalf of the users they run under. That same power is also what makes them dangerous: agents act on whatever content they take in, and much of it comes from outside the user's control — a web page, a repository, a command's output. A single malicious instruction hidden in that content can turn an agent against the very environment it's trusted to work in. With access to source...