Anthropic says it will watermark text generated by its AI models

In a move to align with evolving international standards, Anthropic has announced that it will implement watermarking technology across its suite of AI models, including the popular Claude interface. This update, detailed on the company’s official support page, arrives as AI developers scramble to comply with the EU AI Act’s Transparency Code, which officially went into effect on August 2.

Meeting Regulatory Standards

The EU’s new regulatory framework mandates that AI companies must label machine-generated or AI-edited content in a format that remains detectable by other digital systems. To meet these requirements, Anthropic has confirmed that all models launched after the August 2 deadline will feature built-in technology to watermark both text outputs and generated files. For file-based content, the company is adopting the C2PA open standard, a move that signals a broader industry shift toward verifiable provenance.

Persistence and Scope

Anthropic emphasized that its watermarking strategy is designed for durability. Because the watermark is embedded directly into the text, it is intended to remain intact even when users copy and paste content into other applications or perform minor edits.

“Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from,” the company stated.

While the company has not yet clarified the exact threshold of editing required to strip these watermarks, the implementation will span across its entire ecosystem, including:

• Claude web interface
• Claude Platform API
• Claude Code
• Claude Cowork
• Claude Tag

A Growing Industry Trend

Anthropic is far from alone in this endeavor. As regulatory pressure mounts and public discourse regarding "AI-fishing"—the deceptive use of AI-generated content—intensifies, major tech players are rushing to adopt labeling protocols.

Recent industry actions include:

• Suno, the AI music platform, recently announced plans to tag AI-generated tracks following a series of legal challenges.
• Substack has partnered with Pangram to flag AI-generated newsletters.
• Industry giants including Google, Meta, Microsoft, OpenAI, Black Forest Labs, and Synthesia have all committed to the EU’s transparency guidelines.

As the landscape of generative AI matures, these watermarking initiatives represent a critical step toward establishing trust and accountability in the digital information ecosystem.