In response to the EU AI Act, text generated by Grok models will carry an invisible watermark
English text is being prepared.
Starting November 7, SpaceXAI will add an invisible watermark to text generated by Grok models; detectors can identify this signal to determine whether text carries a watermark invisible to the naked eye. Officials say the move is in response to the EU Artificial Intelligence Act. The watermark has been enabled across SpaceXAI APIs in all regions worldwide, covering the reasoning, non-reasoning, and multi-agent versions of grok-4.7, grok-4.6, grok-4.5, grok-4.3, and grok-4.20, and team administrators can disable the watermark for an individual model. The watermark is not a hidden character or metadata, but is embedded in the model's token selection process when generating text.
Why read this
AI score 59
The original text explains the range of models covered by the watermark and the opt-out option, offering insight into how the EU AI Act's requirement to label generated text is being implemented in practice.
Topics
Body
响应欧盟 AI 法案，Grok 模型生成的文本将添加隐形水印
2026-11-01 14:11 IT之家 - 潞源
IT之家 11 月 1 日消息，SpaceXAI 将于 11 月 7 日起为 Grok 模型生成的文本添加隐形水印。检测器通过识别该信号，判断一段文本是否带有肉眼不可见水印。
官方表示，本次调整是为了响应欧盟《人工智能法案》要求，全球所有地区的 SpaceXAI API 均已启用水印，包含 grok-4.7、grok-4.6、grok-4.5、grok-4.3、grok-4.20 reasoning、grok-4.20 non-reasoning 和 grok-4.20 multi-agent 模型，团队管理员可选择为某款模型单独关闭水印。
IT之家注意到，这种隐藏式水印并非通过隐藏字符或元数据实现，而是融入模型生成文本时对 Token 的选择过程。
AI 治理观察者安德鲁 · 库兰（IT之家注：Andrew Curran）则对此表示，目前还没有为文字加入隐形水印的模型仅有北美地区 ChatGPT。