Within four hours of Anthropic confirming that Claude models would embed invisible watermarks into AI-generated content, developer Guillaume Meyer published code to remove them. His tool has since gone viral on GitHub, with over 20,000 bookmarks on X and more than 100 contributors. The code has also been integrated into numerous projects, according to WIRED. 'Anthropic is embedding watermarks in its Claude texts… the issue is practically history just one day later,' wrote one AI specialist, sharing an image of Meyer breaking chains and standing on crumpled EU and Anthropic flags. Meyer and others began investigating watermarking after Anthropic announced last week that Claude would adopt it to comply with the EU’s AI Act. Some users seek to evade the watermarking due to disagreements with mandatory labeling, while others, including Meyer, enjoy the technical challenge. Freelance writers and social media creators have also reached out for assistance using the code, he said.
The EU’s new rules, which took effect earlier this month, require model providers like Anthropic and OpenAI to label synthetic content so it can be detected as AI-generated, or face fines up to 3 percent of annual turnover. While providers cannot market circumvention tools, there is no legal restriction on independent tools. 'I'm not against transparency, and I'm all for content attribution,' says Meyer. 'I just think watermarking in itself is a really bad solution, because it has major drawbacks and risks.' He is concerned about the risk of false positives and that the watermarking might not distinguish between light or heavy AI use, especially since he often uses Claude and other tools like Grammarly to edit his writing. Using the watermark as evidence—when even Anthropic admits it can only generate a probability that the text has been touched by Claude—could lead to unfair accusations, he says.
Anthropic watermarks text invisibly by leaving a pattern in Claude’s choice of words and phrases that is indiscernible to a human reader but detectable by a machine. Some users are concerned this may degrade the quality of Claude’s responses, though Anthropic insists this won’t be the case. The technique, called SynthID, was developed by Google, which has used it since 2023. Computer scientist Scott Aaronson proposed a similar method at OpenAI but says the firm never deployed it due to customer concerns. Meyer’s removal method uses a non-watermarking large language model to generate multiple rewrites, swapping in synonyms and slightly reorganizing content. This relies on using other models that do not insert watermarks, though 190 organizations, including OpenAI, Microsoft, and Meta, have signed the EU’s transparency code of practice. It remains unclear how many of these labs will implement watermarks, which must be included in all new models from August and integrated into existing models by December.
Source: wired
