“Uses alarm-inducing framing ('dangerous,' 'sues') emphasizing harm and negligence without context on outcome or AI safety measures.”