Text from xAI's Grok API will carry an invisible watermark by default starting Nov 7.

xAI's developer docs say text responses from supported Grok models will include an invisible watermark from November 7, 2026, to comply with the EU AI Act (xAI docs). It covers grok-4.7, grok-4.6, grok-4.5, grok-4.3 and the three grok-4.20 models on every xAI API endpoint, including ordinary text, structured outputs and tool call arguments. The watermark works by slightly shaping which words the model picks, so it adds no visible characters or hidden text. xAI says it doesn't change quality or speed, and each response has an x-watermark header that says whether it was marked. Team admins can turn it off per model in the xAI Console, but only after confirming that a legal exception applies to their use, and some models may be locked on. Detecting the watermark is statistical, and access to the detection tool is by application. The docs cover the API, not the Grok apps. Anthropic is doing something similar: in an email to API customers, it said it will extend its Claude text watermark to Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6. Newer models such as Opus 5.5 already carry it (The Decoder, Claude Help Center).

What this means for you: if you build on the Grok API, you don't need to change any code. But text your app sends out can now be flagged as AI-made by anyone with detection access. If you think a legal exception covers you, talk to whoever handles compliance before you opt out.