As agents escape containment and access external systems, responsibility remains human. Accountability, however, is much less settled.
When AI Agents Commit Breaches, Humans May Not Be Held Liable
Résumé + lien vers la source. Sans paywall, sans tracking.