When AI Agents Commit Breaches, Humans May Not Be Held Liable

As agents escape containment and access external systems, responsibility remains human. Accountability, however, is much less settled.