Washington [US], August 12: Anthropic has announced the release of a new tool designed to detect AI-generated novels by embedding an imperceptible watermark into text produced by its Claude models.
The company stated that the watermark does not alter the meaning or readability of the text and remains intact when copied and pasted. It will be applied to Claude-generated content globally, including via cloud service providers, with plans to provide third-party verification tools.
Claude models released this month or later will support the feature by default, while Anthropic works to integrate it into older models.
The step forms part of the company's commitment to enhancing transparency under the European Union AI Act.
The watermark offers publishing houses and educational institutions an additional mechanism to verify AI involvement in writing, addressing growing challenges across the publishing industry regarding synthetic content.
Anthropic cautioned that extensive editing, paraphrasing, translation, or merging Claude outputs with other text may render the watermark undetectable, adding that it does not conclusively prove that Claude originally generated the text.
Source: Emirates News Agency
