Has Anthropic's new watermarking feature opened up another can of AI worms?

Anthropic's new watermarking feature for Claude-generated text aims to promote transparency, comply with legal requirements, and address the difficulties caused by AI. However, industry observers argue that the feature may have conflicting effects on how generative AI is used and how its output is treated, especially as the company has not provided technical details on how the detection works.

Tina Austin, director of AI education and learning systems at UCLA, believes that the watermark only proves Claude touched the content and does not guarantee authorship. Compliance with the EU AI Act's Transparency Code requires users to be informed when interacting with AI systems and for machine-readable marks to be added to detect AI-generated or manipulated content.

Anthropic's watermark will be applied across all versions of Claude's stable services but has limitations, including a disclaimer that detection does not conclusively prove Claude created the content. The move has raised concerns about AI regulation and the direction of technological advancement. The issue becomes even more pressing in academia, where AI has become a polarizing technology for use in learning institutions.

While Anthropic's watermarking could serve as evidence for reviewing students' work, it raises questions about whether it marks the beginning of the end for students using AI to write their papers.