I was testing this autocomplete website that's based on the Machine Learning model GPT2 (that uses Transformers). I sent as an input to be autocompleted: "Transformers models are not conscious, ". Its first response was "But they can still get hurt".