From a policy perspective, it would only be appropriate that any company who creates a machine capable of large-scale harm should be held liable for neglecting

From a policy perspective, it would only be appropriate that any company who creates a machine capable of large-scale harm should be held liable for neglecting to implement hard-coded safety restrictions. It should not matter whether the machine "thinks" or "feels", for it was created by citizens nevertheless, who have an obligation just like an automobile manufacturer not to do negligent or incompetent things. Likewise, a thinking machine could be seen as something of a pet, and owners of pets (such as dogs, tigers, or gorillas) are certainly liable for not protecting the public appropriately.