Anthropic Claims New System Can Detect AI-Written Text

A new digital battleground is emerging in the fight against AI-generated text. Anthropic, the creator of the Claude chatbot, announced on August 10 the release of new systems designed to digitally watermark content produced by its AI. This move comes in response to new European Union regulations, effective from August 2, which mandate that major AI companies must label AI-generated content.

Anthropic asserts that its new watermarking system is effective worldwide, not just within Europe. The technology aims to embed invisible digital labels within the AI's output, allowing for the identification of text created by Claude. This development is seen as a significant step in addressing concerns about the proliferation of AI-generated content, particularly in academic and professional writing.

However, the practical reality of detecting AI-written text is likely to be more complex. Existing detection systems vary in reliability, sometimes producing absurd results while occasionally offering accurate indications. These services are constantly challenged by increasingly sophisticated AI models like ChatGPT, Claude, and Gemini, whose outputs are often edited by users to appear more 'human.' The effectiveness of Anthropic's new watermarking system against such advanced and potentially re-edited AI content remains to be seen.