AI Slop Identified: Anthropic Adds Invisible Watermark to Claude’s Text Output

Business Insider reports that the watermark does not change the meaning or readability of the text, according to Anthropic. It travels with the content when copied and pasted and “may persist through some editing,” the company said.

A watermark is a faint pattern or hidden marker embedded in digital content, such as photos or documents, to help identify its origin and prevent unauthorized copying.

Anthropic said the feature is part of its commitments toward greater transparency under the European Union AI Act. Upcoming Claude models will support the watermarking from the start, and Anthropic said it is working to extend the function to older models. The markings will apply to Claude-generated content worldwide, including when the model is accessed through cloud providers. Anthropic also said it plans to give third parties tools to detect the watermarks.

Read Full Article Here