Anthropic宣布已签署欧盟AI法案第50(2)条关于AI生成内容透明度的行为准则，并将在Claude生成的内容中采用多种技术标记方式[1]。该公司将通过在文本中嵌入不可见水印和在文件中附加签名来源元数据两种方式，标记所有Claude产品生成的内容[1]。这些水印可随文本的复制粘贴传播，在某些编辑后也能保留，而签名元数据则遵循Coalition for Content Provenance and Authenticity（C2PA）开放标准[1]。
2026年8月2日及之后发布的新Claude模型将从上线之日起支持机器可读标记，此前发布的模型也在逐步添加标记支持[1]。这一标记机制将应用于Claude的所有产品和平台，包括API、Claude Platform、Claude Code、Claude Cowork以及Claude Tag等[1]。
Anthropic has committed to implementing transparent labeling of AI-generated content by signing the European Union's Code of Practice on Transparency of AI-Generated Content under Article 50(2) of the EU AI Act [1]. The company will deploy two complementary technologies to mark text produced by Claude: invisible watermarks embedded within the generated text, and signed source metadata attached to files [1].
New Claude models launched on or after August 2, 2026 will support machine-readable marking from their release date, while previously deployed models are receiving gradual updates to enable marking capabilities [1]. The watermarks are designed to persist through copying and pasting, and can remain detectable even after certain edits [1]. The signed metadata will adhere to the Coalition for Content Provenance and Authenticity (C2PA) open standard [1]. These marking mechanisms will be applied uniformly across all Claude products and platforms, including the Claude Platform, API, Claude Code, Claude Cowork, and Claude Tag [1].