In a major move for AI transparency, Anthropic has announced that all text generated by its Claude models will now carry an invisible, persistent watermark. The change applies globally across every Claude product and is designed to make AI-written content easier to identify—even after it’s been copied, pasted, or lightly edited.

What’s Changing?

Starting with models released on or after August 2, 2026, Claude embeds a statistical watermark directly into the text it generates. Unlike metadata tags that can be stripped away, this watermark is part of the text itself. That means it travels with the content wherever it goes, while remaining completely imperceptible to human readers and without affecting readability.

The rollout covers all Claude surfaces worldwide, including:

  • claude.ai
  • The Claude API
  • Claude Code, Claude Cowork, and Claude Tag
  • Cloud partner deployments on AWS, Google Cloud, and Microsoft Foundry

In addition to text, supported file formats like PNG, JPG, and SVG will include digitally signed provenance metadata where possible.

Why Is Anthropic Doing This?

The initiative stems from Anthropic’s commitment to the European Union’s AI Act, specifically the Article 50(2) Code of Practice on Transparency of AI-Generated Content, which took effect on August 2, 2026. Although the regulation is EU-focused, Anthropic is applying the watermarking standard globally across all its products.

What This Means for Users

For educators, journalists, platforms, and everyday users, the watermark offers a new tool to detect undisclosed AI use. It could help curb academic cheating, reduce misinformation, and promote clearer labeling of AI-assisted content.

That said, the technology isn’t foolproof. Heavy rewriting, translation, or deliberate obfuscation can weaken or remove the signal, and detecting the watermark still requires specialized tools. So while it’s a significant step forward, it’s one piece of a broader transparency ecosystem—not a standalone fix.

The Bigger Picture

Anthropic is the first major frontier AI lab to deploy production-scale text watermarking across its entire product line. The move puts pressure on competitors to follow suit and signals a shift toward built-in accountability in generative AI.

As AI becomes more embedded in daily workflows, features like invisible watermarking may soon become the norm—helping users distinguish between human and machine-created content with greater confidence.

For more details:

https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content