Enterprises deploying AI agents risk compromise when attackers exploit vulnerabilities in legacy infrastructure that these agents depend on, such as unpatched servers, misconfigured Active Directory permissions, and excessive cloud access privileges. Security programs must adopt an exposure management approach that maps and secures the entire attack path—from network and identity layers through cloud infrastructure to AI agent resources—to prevent attackers from leveraging inherited permissions and legacy exposures to hijack AI agents.
https://thehackernews.com/2026/06/stop-your-legacy-infrastructure-from.html