Explained: Anthropic’s plan to watermark all Claude-generated content, and how it works

In a bid to make AI-generated material easier to identify, Anthropic has said it is working to add machine-readable marks to Claude-generated content.

The move comes after the European Union's AI Act took effect. Anthropic has signed the Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content, a voluntary framework aimed at making AI-generated material easier to identify.

How it will work

According to a post on Anthropic’s website, Claude models launched in the EU on or after August 2, 2026 will support machine-readable marking at launch, with generated text carrying embedded watermarks and generated files having digitally signed provenance metadata where supported.

Invisible watermarks will be woven directly into text generated by Claude models. It will not alter the meaning, quality, or readability of a response and can travel with the text when copied elsewhere, potentially persisting through some editing.

For files generated via Claude, supported file types such as .svg, .png, or .jpg, will include signed provenance metadata. If a signed metadata label is present, it signals that a file was processed by Claude and lets you detect whether the file has been tampered with.

Applicable to full Claude suite
Anthropic said Claude markings will cover output from supported models, including Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag.

Embedded watermarks will apply to all generated text. Provenance metadata will apply where Claude supports processing files.

Conditions apply
The AI giant cautioned that a detected mark does not mean it is a proof of AI authorship. “Claude may not be the original author. People often use Claude to proofread, translate, summarize, or convert files. The output can carry a Claude mark even if the underlying ideas, text, or data originated from another source,” it said.

Similarly, the absence of a mark also does not confirm that the content wasn't AI-generated, since older models, heavy editing, very short passages, or metadata stripped through format conversion could all prevent detection.

EU’s AI transparency rule in focus
The European Union's new artificial intelligence transparency rules require companies to clearly label AI-generated content, including deepfakes, to help users distinguish between authentic and synthetic material.

Under the new rules, companies operating AI systems, such as chatbots, must inform users that they are interacting with artificial intelligence. Text, images and other content generated using AI must also carry clear labels, which may include watermarks or other markers to facilitate identification.

Existing AI systems have until December 2 to adapt to the new rules, and there are exemptions for "artistic, creative, satirical, fictional" work.