Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding tasks

Grok API watermarks text from today

Written November 7, 2026

From November 7, seven Grok models watermark API text by default, citing the EU AI Act. What the mark is, how to check it, and the opt-out catch.

What starts today

Starting November 7, 2026, text responses from seven Grok models through the SpaceXSI API carry an invisible watermark. It is on by default, and you can opt out per model. The seven models are grok-4.7, grok-4.6, grok-4.5, grok-4.3, grok-4.20-reasoning, grok-4.20-non-reasoning, and grok-4.20-multi-agent. SpaceXSI says it watermarks text output to comply with the EU AI Act. There is no request parameter to send, and an integration works the same way with or without the mark.

The watermark is invisible. It is applied while the model generates text, slightly shaping which tokens the model picks, so it adds no characters, markup, or hidden text to the response. Detection is statistical: SpaceXSI scores a piece of text and reports whether it is watermarked, not watermarked, or uncertain, and longer text gives a stronger signal. Access to watermark detection is granted by application.

What it covers

Watermarking applies to text output from text generation requests, for each model that has it turned on for your team. It covers every text generation API, including the Responses API and the Chat Completions API, at every endpoint including regional ones. It covers ordinary text responses, structured outputs, and tool-call arguments. The docs page covers API output; it says nothing about the Grok app, X, or grok.com, and gives no date for those surfaces.

How to check a response

Read the x-watermark header on the Responses API or Chat Completions API response. x-watermark: true means the output for that request was generated with the watermark. Any other value, or no header at all, means it was not. The docs give a short example with the Vercel AI SDK:

const { text, response } = await generateText({
  model: xai.responses('grok-4.7'),
  prompt: 'Write a haiku about the ocean.',
});
console.log(response.headers?.['x-watermark']);
console.log(text);

If you ship Grok output to users, the header is the cheap check to add first: log it alongside the request, and you can say for any given response whether the mark was applied, instead of guessing from the model name or the date.

Opting out

Watermarking is controlled per model for your whole team, and only team admins can change it. In the SpaceXSI Console, the Model output watermarking row shows On (all models), Off (all models), or Custom. An admin clears the checkbox for each model to opt out, then saves and confirms with Opt Out. The setting applies to every API key on the team, and changes apply to new requests within 5 minutes.

The catch is in the confirmation. Opting out asks the admin to confirm that an exception under applicable law applies to the use, and that the customer alone is responsible for any marking, disclosure, or other legal requirement. Some models may be locked on for a team: locked models show in the dialog without a checkbox and cannot be turned off from the Console. Every opt-in and opt-out is recorded in the team's audit log.

What to do

Limits

Related playbooks

Sources

Back to: Coding playbooks hub · Also: Claude's text watermark, and the temperature gap · Claude Code's advisor models: Sonnet for speed, Opus for depth