Isn’t this a similar issue with autonomous vehicles? Even if autonomous vehicles reduce overall harm (or are expected to do so), we tend to hold autonomous systems to a much higher standard than humans, which suggests we may also judge chatbots more harshly in this regard.