Anthropic to Add Invisible Watermarks to Claude Text Worldwide
August 12, 2026
Anthropic will embed invisible watermarks in text generated by Claude, marking AI output in a way machines can detect.
The move responds to the EU AI Act’s transparency obligations, which began applying on August 2, 2026 — and Anthropic is applying the Claude watermark for supported models everywhere Claude is offered, not just in Europe.
The decision has drawn pushback from some users, along with supportive reactions.
How the Invisible Claude Watermark Works
According to Anthropic’s official support documentation, supported Claude models weave an imperceptible watermark directly into generated text without changing its meaning, quality, or readability.
The watermark travels with text when it is copied and pasted and may persist through some editing. Supported generated file types, such as SVG, PNG, and JPG, carry digitally signed provenance metadata based on the C2PA open standard.
The marking applies across the Claude API, the Claude apps, Claude Code, Cowork, and Tag, as well as access through AWS, Google Cloud, and Microsoft Foundry. Anthropic notes that signed provenance metadata may not be supported on every platform.
The EU Transparency Obligations Behind the Worldwide Rollout
The change stems from the EU AI Act’s Article 50 transparency obligations, which began applying on August 2, 2026 and require providers of generative AI systems to ensure that AI-generated or manipulated outputs are marked in a machine-readable format and detectable as such.
To support compliance, Anthropic has signed the Code of Practice on Transparency of AI-Generated Content, a voluntary code developed by independent experts through a multi-stakeholder process facilitated by the European Commission’s AI Office.
According to the European Commission’s list of signatories, Google, Meta, Microsoft, and OpenAI have also signed the code. Violating the Article 50 transparency obligations can draw fines of up to 15 million euros or 3% of global annual turnover, whichever is higher, under Article 99 of the AI Act.
Anthropic says Claude models launched on or after August 2, 2026 support marking at launch, and that it is working to add support to earlier models. The company applies the policy everywhere Claude is offered.
A Detected Mark Is Not Proof of Authorship — and Some Users Object
TechCrunch highlighted several Reddit posts that called the watermarks unethical, arguing that the users supplied the instructions and judgment while Claude served as a tool. The same article also highlighted posts supporting the watermarks and pushing back on the critics; these posts are not a survey of overall user sentiment.
Anthropic is explicit about the limits. A detected Claude watermark indicates content may have been processed by Claude, not that Claude authored it. Summarizing, translating, or proofreading can leave a mark on largely human work, while heavy editing, paraphrasing, or short passages can make the watermark undetectable.
For Anthropic, which has confidentially filed for an IPO, the rollout tests how the company balances regulatory compliance with user trust.
What to Watch Next
Anthropic has not said when models released before August 2 will gain watermark support, and it remains unclear how much editing strips the mark.
The company plans to publish technical documentation so users and third parties can detect the marks, and those details will determine how useful detection is in practice.
Sources
Anthropic – How Claude marks AI-generated content
TechCrunch – Anthropic says it will watermark text generated by its AI models
European Commission – Code of Practice on Transparency of AI-generated Content
EUR-Lex – Regulation (EU) 2024/1689 (Artificial Intelligence Act)
