dil-leik-og comments on Alignment Faking in Large Language Models