Authorship and the Politics and Ethics of LLM Watermarks

Tim Räz

Recently, watermarking schemes for large language models (LLMs) have been proposed to distinguish text generated by machines and by humans. The present paper explores philosophical…