[bkmeneguello]

I think it's because it uses the mainstream media as "reliable" dataset. There was another AI that was trained on Reddit that was considered toxic and with "far right" opinions.