xAI to add invisible watermarks to Grok-generated text from November 7 across all regional APIs
SpaceXAI will begin adding invisible watermarks to text generated by Grok models on November 7, with a detector identifying the signal to judge whether a passage carries an imperceptible watermark.
SpaceXAI will begin adding invisible watermarks to text generated by Grok models on November 7, with a detector identifying the signal to judge whether a passage carries an imperceptible watermark. The company said the change responds to the EU AI Act and that watermarking is already enabled for SpaceXAI APIs in all regions, covering 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, with team administrators able to disable watermarking for a specific model. The hidden watermark is not implemented through hidden characters or metadata but is woven into the model's token selection while generating text. AI governance observer Andrew Curran noted that ChatGPT in North America is currently the only major model without invisible watermarking on text.