It’s about to get much easier to spot AI-generated text, with Anthropic announcing that invisible AI watermarks will soon be added to all Claude-created content.
The new policy, which is mandated by the European Union’s AI Act, will apply to any text or files that were created or otherwise touched by Claude, including computer code, company reports, or cover letters to potential employers.
In a recently published support page, Anthropic says that new Claude models launched in the EU on or after August 2, 2026 will support adding watermarks to the content they create or edit, and that it’s working to add watermarking abilities to previously launched Claude models.
The watermarking mandate stems from the EU AI Act, which includes a provision boosting the transparency of content that was created or “processed” by AI.
Anthropic says it will deploy Claude’s watermarking abilities worldwide and across all Claude products, including Claude Cowork, Claude Code, and the Claude API, to comply with the law.
When it comes to marking text, Anthropic says a “supported” Claude model “weaves an imperceptible watermark into the text itself,” and that the mark will remain present after the text is copied and pasted, or even after “some” editing.
Meanwhile, Claude will attach “provenance” metadata to certain files it creates, including SVG, PNG, and JPG files.
Anthropic says it plans on releasing tools that allow anyone to detect the watermarks, with details on those tools slated to arrive in “forthcoming” announcements.
The presence of a Claude watermark doesn’t necessarily mean the content was originally created by Claude, Anthropic points out. For example, human-written text that’s later edited by Claude will also get a watermark.
There may also be cases where Claude-touched content isn’t watermarked, such as when the content is “heavily” edited, too short, or produced by a Claude model that doesn’t yet support watermarking.
It’s worth noting that the EU AI Act applies to any AI providers operating in the European Union.
Both Google and Meta have pledged to sign the act, while ChatGPT maker OpenAI says it will take a “layered approach” to the act’s transparency provisions. But while those companies have signaled their intentions to comply with the EU AI Act, they’ve stopped short of detailing their own watermarking plans for AI-generated text.
In any event, news of Claude’s forthcoming watermarks means a) it will soon get much easier to detect Claude-created or “processed” text, and b) it will also be much easier for others to spot text that you’ve created with Claude. Food for thought.
