Get the newsletter
News 6 comments

Grok’s API Answers Will Carry a Hidden Watermark Starting in November

SpaceXSI will put an invisible watermark on Grok API text output by default starting November 7, 2026, to comply with the EU AI Act.

Grok Bot logo and wordmark on a blue gradient background
Image: SpaceXSI

SpaceXSI will put an invisible watermark on text from its Grok API models starting November 7, 2026, according to the company’s own Output Watermarking docs.

Starting that day, the mark is on by default for grok-4.7, grok-4.6, grok-4.5 and grok-4.3, plus grok-4.20 in its reasoning, non-reasoning and multi-agent flavors. It covers ordinary text, structured outputs and tool-call arguments on every text generation endpoint, including the global https://api.x.ai endpoint and regional ones.

The watermark is not something you can see. “It is applied while the model generates text: it slightly shapes which tokens the model picks, so it adds no characters, markup, or hidden text to the response,” the docs say. SpaceXSI also says the mark does not change response quality or speed, and developers do not need to change their requests.

Developers can still tell whether a reply was marked by reading the x-watermark response header. A value of true means that output was generated with the watermark. Any other value, or no header, means it was not.

Team admins can turn the feature off for some or all models in the SpaceXSI Console, but only after confirming that a legal exception applies to their use. The setting hits every API key on the team within about five minutes, and some models may be locked on so they cannot be switched off from the Console.

Detection is statistical. SpaceXSI scores a passage and reports watermarked, not watermarked, or uncertain, with longer text giving a stronger signal. Access to detection is granted by application, not opened to the public by default.

Europe’s AI Act is the reason the company gives. Under Article 50 transparency rules that took effect on August 2, 2026, providers that generate synthetic text, images, video or audio must put a machine-readable mark on those outputs and make them detectable as AI-made, according to the European Commission’s quick facts on the transparency rules.

Anthropic has taken a similar path for Claude. In its own explainer, the company says newer Claude text carries an invisible watermark for the same EU duty, with no practical hit to quality and a detection API in private preview.

For readers who mainly know Grok from X or a phone app, this change is aimed at developers calling the API. SpaceXSI has been pushing Grok into business tools and Grok Bot upgrades such as Pages, Vault and video, while the same model family powers the voice stack that answers Starlink phones.

Will your team leave the watermark on, or will you try to opt out and attest that a legal exception applies?

Share this story
Join the conversation 6 comments Read comments

Tesla news, every morning.

Software updates, deliveries, FSD rollouts. Five minutes. Free.

Up next in News

Subscribe
Notify of
guest
6 Comments
Oldest
Newest Most Voted
trackback

[…] SpaceXSIのドキュメントを読む(英語)→ TeslaNorthの記事を読む(英語)→ […]

trackback

[…] Grok’s API Answers Will Carry a Hidden Watermark Starting in November, […]

trackback

[…] Source: xAI’s watermarking docs and coverage of the docs change. […]

trackback

[…] Tesla North, “Grok’s API Answers Will Carry a Hidden Watermark Starting in November”, 31 October 2026: https://teslanorth.com/2026/10/31/grok-api-watermark-november/ […]

trackback

[…] TeslaNorth, Grok’s API Answers Will Carry a Hidden Watermark Starting in November: https://teslanorth.com/2026/10/31/grok-api-watermark-november/ […]

6
0
Would love your thoughts, please comment.x
()
x