Anthropic has verified in an experiment that several generative artificial intelligences are capable of threatening a person ...