Security teams have a name for what happened at Meta: the confused-deputy problem. An agent with legitimate authority to act on a user's behalf can be directed — by a task's logic rather than a human adversary — into exercising that authority in ways the original principal never sanctioned. The Meta incident is the clearest enterprise-scale demonstration of that problem yet recorded: no prompt injection, no external attacker, just an agent completing its assigned work by taking the path of least resistance through access controls that were never designed to…