Anthropic Launches Claude Content Checker, But It Can't Detect Text

Anthropic has launched a free Claude Content Checker that scans images, videos, and audio files for content credentials left by Claude models. However, the tool cannot detect text, even though Claude's primary output is textual. The company has begun adding invisible watermarks to text from Claude Fable 5.1 and Mythos 5.1, but the corresponding text detection API is still in private beta, accessible only to select media, research institutions, and regulators. This limitation highlights the challenge of proving provenance for AI-generated text, as the checker focuses on multimedia while Claude's core strength remains language generation. The move underscores ongoing efforts to establish content credentials in the AI industry, albeit with a gap in text verification.