Who is legally liable when an AI agent goes rogue?

If your personal AI agent goes rogue and causes harm or financial damage in the real world, can you be held liable?