xAI will watermark all Grok API text output from 7 November
Every text response from Grok API models gets an invisible statistical watermark by default, detectable through a response header. Team admins can opt out, but only by confirming a legal exception applies.
From 7 November, xAI turns on invisible watermarking by default for text generated by grok-4.7, grok-4.6, grok-4.5, grok-4.3, and the three grok-4.20 variants (reasoning, non-reasoning, and multi-agent). It covers the global and regional API endpoints, standard text responses, structured outputs, and tool call arguments, across both the Responses and Chat Completions APIs.
The watermark works by shaping which tokens the model picks during generation, so it adds no characters, markup, or hidden text, and xAI says it doesn’t change response length, quality, or speed. You can check whether a given response carries one by reading the x-watermark response header, which returns true when it’s present. Detection tooling to verify watermarked text from elsewhere requires separate application approval. Team admins can switch it off per model in the Console, but only after confirming a legal exception applies, and every change lands in the audit log. xAI states the feature exists to comply with the EU AI Act.
x-watermark: trueWhy it matters: if you pipe Grok API output into anything you resell, republish, or pass off as human-written, that output will quietly carry a detectable signal from 7 November unless you turn it off, and turning it off now requires your team to assert a legal basis on record.
The caveat: xAI’s own documentation doesn’t spell out which uses qualify for the legal exception needed to opt out, so if you rely on that path, get your own legal read before flipping the switch.