Skip to main content
Back to News Hub
🟢TechCrunch AI
August 11, 2026
Claude

Anthropic says it will watermark text generated by its AI models

Overview

Anthropic announced that it will implement text watermarking across its AI models, including Claude, to comply with European Union regulations. The watermarking technology automatically applies to content produced by models released after August 2, with plans to support older models as well. Because this system operates at the model level, the identifying marks remain attached to text when users copy and paste it into other locations.

Key Takeaways

  • To align with the EU AI Act's Transparency Code that came into effect on August 2, Anthropic updated its support page to confirm that text generated by its Claude models will now feature watermarking.

    Models launched after August 2 will automatically include technology to watermark both generated text and files, utilizing the C2PA open standard for file formats.

  • The company also intends to expand this watermarking capability to legacy models, applying the indicator at the foundational model level across products including Claude Code, Claude Cowork, Claude Tag, and the Claude platform API.

    Because this watermarking functions at the model layer, the embedded identifiers travel directly with the text when individuals copy and paste it across different platforms.

  • While it remains unclear how much manual editing is required to strip the watermark, Anthropic's move reflects a broader industry shift toward compliance with European mandates.

    Other tech entities, including OpenAI, Google, Meta, Microsoft, and Black Forest Labs, have similarly pledged adherence to the EU code, alongside platforms like Suno and Substack taking steps to address AI-generated content.

  • Anthropic is adding built-in text watermarking to its AI models to comply with the EU AI Act's Transparency Code that took effect on August 2.

    The watermark operates at the model level across surfaces such as the Claude platform API, Claude Code, Claude Cowork, and Claude Tag.

  • For file generation, Anthropic is adopting the C2PA open standard to mark computer-generated content.

To align with the EU AI Act's Transparency Code that came into effect on August 2, Anthropic updated its support page to confirm that text generated by its Claude models will now feature watermarking. Models launched after August 2 will automatically include technology to watermark both generated text and files, utilizing the C2PA open standard for file formats. The company also intends to expand this watermarking capability to legacy models, applying the indicator at the foundational model level across products including Claude Code, Claude Cowork, Claude Tag, and the Claude platform API.

Because this watermarking functions at the model layer, the embedded identifiers travel directly with the text when individuals copy and paste it across different platforms. While it remains unclear how much manual editing is required to strip the watermark, Anthropic's move reflects a broader industry shift toward compliance with European mandates. Other tech entities, including OpenAI, Google, Meta, Microsoft, and Black Forest Labs, have similarly pledged adherence to the EU code, alongside platforms like Suno and Substack taking steps to address AI-generated content.

Anthropic is adding built-in text watermarking to its AI models to comply with the EU AI Act's Transparency Code that took effect on August 2. The watermark operates at the model level across surfaces such as the Claude platform API, Claude Code, Claude Cowork, and Claude Tag. Because the mark is integrated directly into the generated text, it travels alongside the content when users copy and paste it into other applications.

For more details please read the original article at TechCrunch AI.

Continue Learning

Comments

Comments appear only after moderation. Your email identifies your submission to the moderator and is never displayed here.

No approved comments yet.

Originally published by TechCrunch AI
Read the original