Anthropic will embed invisible watermarks into text generated by its Claude AI models and digitally sign image files, a move that aligns with new European Union regulations on AI transparency. The company announced that new Claude models launched on or after August 2, 2026, will support machine-readable marking from their release. Existing models will receive this capability during a transition period.
This initiative stems from Anthropic signing the EU AI Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content. While prompted by European regulations, the watermarking will be applied worldwide to all supported Claude products and platforms. This includes outputs from Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag, as well as content accessed through cloud partners like Amazon Web Services, Google Cloud, and Microsoft Foundry.
For text, Anthropic will weave an imperceptible watermark directly into the output. This mark is designed to be invisible to human readers and not alter the meaning, quality, or readability of the response. The watermark is intended to travel with the text when it is copied and pasted and may persist through some editing. However, Anthropic cautions that heavy editing, paraphrasing, translation, or the generation of very short passages could make the mark undetectable.
Supported file types, such as .svg, .png, and .jpg images, will receive digitally signed provenance metadata adhering to the Coalition for Content Provenance and Authenticity (C2PA) standard. This metadata provides a digital signature indicating that a file was processed by Claude and can help detect tampering.
Anthropic emphasizes that the presence of a watermark or metadata signal does not serve as definitive proof of AI authorship. The company noted that content processed by Claude, even if originally human-written and used for tasks like proofreading, translation, or summarization, could carry the mark. Conversely, content generated by older models or that has undergone significant alterations might not carry a detectable mark.
Technical details regarding the watermarking algorithm and detection mechanisms are not yet fully published, though Anthropic plans to release further guidance. The announcement has drawn varied reactions online, with some users expressing concern about the implications for content creators and coders, while others see it as a necessary step for AI transparency.