Anthropic's Claude AI Now Has a Content Checker

Anthropic has launched a new web-based tool, accessible at claude.com/check-content, designed to identify content processed by its Claude AI. This checker operates by examining embedded content credentials, specifically adhering to the industry standard known as C2PA.

Initially, the checker is limited to validating multimedia files such as JPEGs, GIFs, MP3s, and MOVs. It does not support text-based file formats. Anthropic has indicated that a more expansive AI detection tool for text is in the works and is currently in a private preview phase, with further details anticipated in the coming months.

The company states that the watermarking of AI-generated text will not impact its quality, speed, or cost. These embedded watermarks will persist even when content is copied and pasted. Anthropic is offering businesses and organizations the opportunity to sign up for early access to its watermark detector.

Regarding user privacy, Anthropic assures that uploaded files are not stored or used for any other purpose. The checker only accesses the file's credentials. In testing, the tool successfully identified an image file processed by Claude, while also noting that detection signals can be lost if files are edited or converted.