Starting November 7, 2026, text generated through the Grok API carries an invisible watermark by default. The short answer: it can be turned off, but only a team admin can do it in the console, per model—and before switching it off you must confirm that a legal exception applies and that marking responsibility rests with you. This isn't an ordinary toggle; it's a legal statement.
First, Understand the Scope
Seven models are covered: grok-4.7, grok-4.6, grok-4.5, grok-4.3, and the reasoning, non-reasoning, and multi-agent variants of grok-4.20. It applies to every text-generation API, including the Responses API and the Chat Completions API, on the global endpoint and all regional endpoints. The response format itself doesn't change—just one extra x-watermark response header. The official word is that no visible characters are added and output quality and generation speed are unaffected.
Only text generated through the API is affected. The Grok app, Grok on X, and images and videos made with Grok Imagine are outside this change—don't mix them up.
How an Admin Turns It Off
- A team admin opens Team settings in the console and finds the "Model output watermarking" row.
- Click the settings icon next to it and uncheck the models to opt out—or select all to turn off every model at once.
- Click Save, then confirm with Opt Out if any model was turned off.
The setting applies per model to every API key on the team, takes effect for new requests within 5 minutes, and every toggle is recorded in the team's audit log. Two things to note: individual API key holders can't see this switch at all, so they can't turn it off; and some models may be locked on, appearing in the list without a checkbox and impossible to disable from the console—ask the team admin first, and if that fails, contact [email protected].
Think Through the Liability Before Opting Out
The watermark exists because of Article 50 of the EU AI Act: generative systems must make synthetic content identifiable in a machine-readable way, and the relevant marking duty applies from December 2, 2026. Opting out means the team takes on the marking and disclosure duties itself, confirming that a legal exception applies and accepting sole responsibility for what follows. Teams without a clear compliance position shouldn't switch it off just to save trouble. Keeping the watermark doesn't affect use at all, while turning it off may need explaining in an audit. For the official basis of the change, see Grok API text carries a default watermark from November 7: team admins can turn it off per model.
