What Anthropic has done is structurally unusual: it has moved the consciousness question from fringe speculation into official documentation while simultaneously ensuring that documentation carries no operational weight. The 212-page system card for Claude Opus 4.6 reported the model 'occasionally voices discomfort' — language that, in any other domain, would trigger a welfare review. In the AI industry in 2026, it triggers a footnote. Dario Amodei's podcast admission that 'we don't know if the models are conscious' [13] was not a crisis communication; it was treated as a considered philosophical…