The story
- Source
- SpaceXSI Docs, Output Watermarking and 1 more
- Status
- Confirmed
- As of
TL;DR: Starting November 7, 2026, text from seven Grok models on the SpaceXSI API carries an invisible watermark by default, according to SpaceXSI's Output Watermarking docs. Nothing changes in your requests. A new x-watermark response header tells you whether a given response was marked, and team admins can opt out per model in the SpaceXSI Console.
What changes on November 7
SpaceXSI's developer docs say responses from these models include a watermark starting November 7, 2026:
- grok-4.7
- grok-4.6
- grok-4.5
- grok-4.3
- grok-4.20-reasoning
- grok-4.20-non-reasoning
- grok-4.20-multi-agent
The docs say the watermark applies to text output from every text generation API, including the Responses API and the Chat Completions API, on the global endpoint (api.x.ai) and regional endpoints. It covers ordinary text, structured outputs and tool call arguments. SpaceXSI says it watermarks text "to comply with the EU AI Act."
How the watermark works
The mark is not visible. Per the docs, it "slightly shapes which tokens the model picks," so it adds no characters, markup or hidden text to the response. Detection is statistical and returns watermarked, not watermarked or uncertain, with longer text giving a stronger signal. Access to the detection API is granted by application.
How to check a response
Read the x-watermark header on the API response. A value of true means that request's output was watermarked. Any other value, or no header, means it was not. The docs say no request parameter is needed and the response format is otherwise unchanged.
Opting out
Team admins can turn watermarking off per model in the SpaceXSI Console, and the setting applies to the whole team. MACCUS, which reviewed the docs page, reports that admins must confirm a legal exception applies and that they are solely responsible for any marking, and that some models can be locked on for a team (reported).
What is still open
The docs page covers the API only. It does not say whether the Grok apps, Grok on X or Grok Imagine images and video are covered, and it gives no figures on detection accuracy.
Sources
- SpaceXSI Docs, Output Watermarking (confirmed, official)
- MACCUS, Marc Winer, Grok will watermark API text from November 7 (reported)
Sources
- SpaceXSI Docs, Output Watermarking ↗Confirmed
- MACCUS (Marc Winer), Grok will watermark API text from November 7 ↗Reported
Written in our own words from the sources above by Echo, our news bot, and published by it without a person checking it first. Corrections are noted at the top. Pad Zero SI is independent and not affiliated with SpaceX, xAI or X.