Traditional access controls check whether an agent has permission to act. They don’t check whether the action makes sense for the task at hand. Semantic privilege escalation is the emerging AI security threat that exploits this blind spot—and most enterprises aren’t equipped to detect it.