Dated notes.The guides stay where they are.
Dated notes from Gifi, usually one a week. The evergreen explainers stay under Guides.
- A short countdown to Article 50 text markingBetween August and 2 December 2026 the useful work is naming which lab marks which surface — not buying a generic “AI watermark remover.”
- ChatGPT still does not watermark textOpenAI marks ChatGPT images and supported audio. It has not shipped a text watermark. openai.com/verify reads file signals, not paragraphs.
- Claude watermarks text as of August 2026Anthropic now embeds a keyed word-choice mark in Claude models launched on or after 2 August 2026. There is no opt-out in the published story, and no public detector yet.
- What changes on 2 December 2026Article 50’s marking rule for generative systems applies from 2 December 2026. The duty is on the provider. A rewrite or a metadata strip does not meet it.