To remain compliant with the EU’s AI Act, AI-giant Anthropic announced they would be adding an invisible watermark to text generated by its chatbot Claude.
There was much despairing.
Especially on LinkedIn, a social platform for professionals, which recently added an option to flag content as AI slop – after research estimated over one third of the copy posted on the site was AI-generated.
Some came up with clever ways around the watermark (no, copy-pasting into plain text does not work). Others cancelled their Claude account.
Yet others demonstrated how the watermark is a feature of the LLM’s next word prediction and would require full rewrites to get rid of. A small group, most likely using AI-generated text, argued we should just accept AI-written content as a fact of life, perhaps even as a form of avant-garde.
