Advertise hereReach developers building with AI coding agents. →
Grok Bot Industry News

From November 7, Grok's API quietly watermarks the text it writes. Here's what changes and how admins can turn it off

xAI's developer docs say that from November 7, 2026, text from grok-4.7, grok-4.6, grok-4.5, grok-4.3 and the grok-4.20 reasoning, non-reasoning and multi-agent models will carry an invisible statistical watermark by default. It covers the Responses and Chat Completions APIs, including structured outputs and tool-call arguments. xAI says it's for EU AI Act transparency rules, needs no code changes and doesn't affect quality or speed. Watermarked responses carry an x-watermark: true header, team admins can opt out per model in the Console, and access to the detection API is by application.

Why it matters

If you ship Grok-written text to customers or must meet EU AI Act rules, you'll want to know it's marked, and decide per model whether to keep it on.

Who it's for

Developers and team admins using the xAI API in products, especially those serving users in the EU.

What it means for you

  1. List which Grok models your app calls and check them against the models in xAI's watermarking doc.
  2. After November 7, log the x-watermark response header to confirm what's marked.
  3. If you need it off for a model, have a team admin change it in Console > Team settings > Model output watermarking.

Applies to: On by default for the listed models from November 7, 2026, at no extra cost. Opt-out is per model in Console > Team settings > Model output watermarking and takes effect within about 5 minutes. Detection is statistical and the detection API is by application only. Checked November 1, 2026.

Share on X

Source: xAI docs on docs.x.ai

Written by AgentStackList from the original post. Checked Nov 1, 2026. Features, plans and availability can change, so confirm on the source before you rely on it.

xai-apiwatermarkeu-ai-actdevelopersgrok

More Grok Bot tricks

All Grok Bot tricks →