Gruber argues that AI text watermarking likely degrades prose quality and, even if it doesn’t, remains objectionable because it operates as an embargoed, unverifiable tag or signal controlled by providers like Anthropic and Google.