If AI systems are seen as advanced to control, companies that build them can’t held liable for the harms they cause