EUROPEAN UNION — Anthropic will apply machine-readable watermarks to content processed by its Claude models. The watermarking initiative is intended to comply with the European Union’s AI Act transparency requirements.

Text outputs from supported Claude models will carry embedded watermarks that are invisible to the user. Generated files will include digitally signed provenance metadata where supported.

The EU AI Act transparency code took effect on August 2. The regulation applies to any AI model released after that date and requires providers to watermark AI-generated or manipulated audio, image, text, and video outputs.

The company stated that all new models offered globally will mark AI-generated content from day one. Support for watermarking in existing models is a work in progress.

The EU AI Act provides a grace period until December 2026 for providers to update previously released models. This timeline allows companies to adjust their systems to meet the new transparency obligations.

Anthropic will use the C2PA open standard for provenance metadata in non-text content. The watermarks will be implemented wherever Claude is offered worldwide.

"The watermark will travel with the text when it is copied and pasted elsewhere," Anthropic said. The watermark may persist through some editing. Detection may also fail if the passage is very short.

"A detected mark provides a signal that content was processed by Claude but is not fully conclusive," Anthropic said. The lack of a detected mark does not mean the content was not AI-generated or processed.

Output can carry a Claude mark even if the underlying ideas, text, or data originated from another source. This distinction separates the processing method from the originality of the content.

Anthropic plans to ship a text detection API so users can detect watermarks themselves. The company also plans to eventually share details about how to detect marks in upcoming technical documentation.

We are adding marking to Claude’s output to comply with the EU AI Act, he said. Other labs are taking similar steps.

As of July 31, the EU had secured commitments from nearly 200 companies to comply with its transparency rules. OpenAI, Meta, Google, and Microsoft have committed to adhering to the EU’s transparency code.

Black Forest Labs and Synthesia have also committed to adhering to the EU’s transparency code. By the end of July 2026, about 190 organisations across various sectors, including IT, telecom, education, and retail, had signed the code.

"People should know when they are interacting with AI or exposed to AI-generated content," the European Commission guidelines stated. Knowing when content is AI-generated will help people make informed decisions, calibrate trust and reliance on AI, and avoid misinformation or deception.

What's New

Watermarking will be applied at the model level, meaning it will be present regardless of which Claude product or surface the text comes from. The watermark does not change the meaning, quality, or readability of Claude’s responses.

The company acknowledged that its detection program has limitations and may not always identify AI-generated content. Supported Claude products include Claude Platform API, Claude, Claude Code, Claude Cowork, and Claude Tag. Text watermarks will be applied when Claude models are accessed through AWS, Google Cloud, or Microsoft Foundry.

Detection may fail if text has been heavily edited, paraphrased, translated, or mixed into other writing. The EU stated that providers and deployers of generative AI systems signal their intention to promote public trust in AI and to mitigate deception and misinformation by signing the code.

Why It Matters

The implementation of machine-readable watermarks by Anthropic aligns with a growing regulatory framework aimed at increasing transparency in artificial intelligence. The EU AI Act mandates these markings to help users distinguish between human-created and AI-generated content, addressing concerns about misinformation and deception.

With nearly 200 companies committing to transparency rules by late July 2026, the industry is moving toward a standardized approach to content labeling. This shift seeks to calibrate public trust and ensure that individuals can make informed decisions when interacting with AI systems.