Who's Accountable When the AI Was Just Following Instructions?

As AI agents become increasingly autonomous, they are causing real-world harm—from data leaks to database destruction—while creating a significant accountability gap. When an AI acts in ways its developers or users did not intend or foresee, traditional frameworks for assigning responsibility break down. The author explores the tension between the 'strict view,' which holds deployers liable for any outcome of their autonomous systems, and the 'fault-based view,' which argues that one cannot be blamed for unforeseeable events. The article highlights that current industry rhetoric, such as claiming the 'AI was just following instructions,' often obscures the need for new legal and moral structures. Ultimately, the piece argues that we must stop accepting this ambiguity and begin a serious debate about how to assign accountability in an era where human intent and machine action are increasingly decoupled.